Location qualification foundation + low-hanging-fruit cleanup (prereq for /authors optimization)

You are being dispatched from flower Brief #242: Location qualification foundation + low-hanging-fruit cleanup (prereq for /authors optimization)

Recall pointer:
- Use recall_brief with id 242 for the full folder if you need provenance.

Target:
- project: lounge (/Users/mikeferrara/Documents/code/lounge)
- branch: feature/authors-optimization
- worktree: not specified
- kind: fresh

Current brief spec:
# Location qualification foundation + low-hanging-fruit cleanup

## Goal
Establish a **precomputed, config-driven "qualified location" set** as the clean foundation the /authors optimization epic (#239) builds on. Replace doc-30's maintained live query-time counts with a boolean flag on `locations`, driven by tunable AppSetting conditions, kept fresh by a scheduled sweep, and reconciled into Meilisearch via Scout — so location search/typeahead query a qualified set, not the raw ~206K rows (~94% junk). Non-destructive: junk is flagged out of discovery, not deleted. **Sequenced FIRST; #239 depends on it.**

## Background / evidence
- `locations` ≈ 206K rows on prod, ~94% junk (`everywhere`, `mf`, one-off typos) per the /authors-optimization design (doc 30 / 00-PRODUCTION-FACTS). A distinct-author floor of 5 would keep ~21.8K.
- Today there is no clean set — typeahead/search surface the raw table. Building proximity/search/favorites on this means every downstream piece inherits the junk.
- Operator principle: clean the foundation before building the layer; a solid config-driven foundation, not an exhaustive cleanup.

## Design — qualified flag + config + sweep + Scout
1. **Flag column** on `locations` — a boolean (name TBD at build, e.g. `is_search_visible`; avoid collision with the author `only_indexed` concept). Indexed. Defaults false; the sweep sets it.
2. **AppSetting-driven qualification conditions** (tunable, not hardcoded):
   - `location_qualify_min_name_length` (default 3 — drops `mf`, keeps `PDX`; allowlist any legit 2-char names if found).
   - `location_qualify_min_posts` and/or `location_qualify_min_authors` (low threshold; final defaults tuned by the low-hanging-fruit pass).
   - `location_qualify_blocklist` (seed with known junk: `everywhere`, `mf`, …; extensible).
   - Optional `city_id IS NOT NULL` / coordinate-presence gate if useful for proximity.
3. **Scheduled sweep** — `lounge:recompute-location-qualification` (or similar), scheduled (default daily, off-peak; cadence configurable). Recomputes the flag for all locations against current conditions. Chunked, idempotent, resumable, on the maintenance worker via **inline scheduler — NOT a queued job** (queue-reference 15s-timeout gotcha).
4. **Rebuild-on-config-change** — an AppSetting observer/event: when any `location_qualify_*` key changes, trigger a re-sweep (or mark dirty for the next run) so the qualified set tracks config.
5. **Scout → Meilisearch reconciliation** — location search/typeahead query the qualified set. Either (a) `Location` uses Scout with `shouldBeSearchable()` gated on the flag (unqualified locations drop out of the Meili index), or (b) a flag filter in the search query — pick whichever matches existing Scout usage here. Reconcile the Meili index off the flag like other models.

## Low-hanging-fruit pass (part of this brief)
- Characterize the junk on prod (read-only): distribution by name length, by post/author count, most-common junk names, obvious non-place strings — via `bin/diag` / the read-only SSH PHP-script pattern.
- Tune the initial condition defaults from that characterization.
- Investigate whether tooling helps identify junk at scale (typo / non-place detection) — optional; note findings, don't over-build.
- **Non-destructive:** sets flags/conditions only; does NOT delete or merge location records (deeper cleanup/merge is a separate future cycle).

## Constraints
- **Discovery only.** The flag gates search/typeahead visibility; `resolveFromSlug` reachability stays untouched (existing links keep working).
- Non-destructive (flag, don't delete/merge).
- Schema add is INSTANT (one boolean col + index, online/off-peak on the memory-tight prod box).
- Sweep/reconcile run inline-scheduled, not queued.

## Acceptance criteria
- `locations` has the qualified flag + index; the sweep populates it from AppSetting conditions.
- Changing a `location_qualify_*` AppSetting triggers/permits a rebuild and the set updates.
- Location search/typeahead return only qualified locations; Meili index reconciled off the flag.
- A junk-characterization summary + tuned default thresholds are recorded.
- Zero deletions/merges of location records; reachability unaffected.

## Relationship
Prereq for #239 (/authors optimization); replaces #239's former C3 minimal floor. Broader app-wide location-search centralization is a separate later round.

## Open detail questions (low-stakes; defaults proposed, tunable post-ship)
- Flag column name (`is_search_visible` vs `is_qualified`) — default `is_search_visible`, final at build.
- Sweep cadence — default daily off-peak, configurable.
- Initial thresholds — min name length 3; min posts/authors tuned from the characterization pass.

Recent/key trace events:
[6] status_change lounge-refine: (no body)
[7] participant_joined lounge-orchestrator: (no body)
[8] status_change lounge-orchestrator: (no body)
[9] dispatched lounge-orchestrator: Dispatch request #127 queued for lounge.
[10] dispatched lounge-orchestrator: Dispatch request #127 claimed and spawned as process #1161.
[11] link_added lounge-orchestrator: (no body)
[12] link_added lounge-orchestrator: (no body)
[13] status_change lounge-orchestrator: (no body)
[14] link_added lounge-orchestrator: (no body)
[15] participant_joined lounge-worker-242: (no body)
[16] note_added lounge-worker-242: **BUILD PLAN — crystallized (pre-restart save, process 1161, worktree feature/authors-optimization, clone DB lounge_feature_authors_optimization). NO files/migrations written yet — clean stable state.**

## Recon facts (verified this session)
- Location "name" column is `loc_string` varchar(125) UNIQUE (NOT `name`). Slug=`slug` varchar(58). Coords `lat`/`lng` decimal(10,7). Also `merged_into_id`, `city_id`/`state_id`/`country_id`. NO count cols, NO bool flags today. Model `Location.php` uses `$guarded=['id','created_at']` (new cols auto-fillable); casts geo_*+json_response.
- `locations` has NO create migration — squashed in database/schema/mysql-schema.sql:321. Add col via new migration.
- Distinct-author + post source of truth = **`location_reddit_user` pivot** (composite PK location_id+reddit_user_id, one row per author; col `posts_count`). author_count=COUNT(*), post_count=SUM(posts_count) per location_id. Maintained by ingest + `lounge:backfill-location-reddit-user-pivot` + daily `ReconcileLocationRedditUserPivot` (console.php ~107). Excludes default location.
- Scout: `Location` uses `Searchable` (line 90). `toSearchableArray()` (271-280) = loc_string,country_id,state_id,city_id,top_level. `shouldBeSearchable()` (282-294) = ScoutIndexing::enabled() && !merged_into_id. NO searchableAs → index name = `locations`. Meili filterableAttributes in config/scout.php:308-313 = country_id,state_id,city_id,top_level. Settings pushed via `lounge:syncMeilisearchSettings --force --model=...`.
- ScoutIndexing::enabled() = AppSettings `scout_indexing_enabled` (default true). Explicit `->searchable()` bypasses shouldBeSearchable → guard bulk pushes with ScoutIndexing::enabled(). Reconcile prior art: collection/model `->searchable()` (BuildRedditPostsIndex:143 `$posts->searchable()`, RecountDirtyTextsJob:96, RebuildSearchIndex:100).
- Search call sites: PUBLIC = LocationAutocomplete.php:19 (`Location::search(...)->take(6)->where('top_level',true)->safeGet()`), LocationsIndex.php:35. ADMIN (must keep FULL-index reach) = AdminLocationSearchPanel.php:81 (**the merge/junk-management tool** — do NOT hide junk from it) + LocationController.php:82. LocationSearch.php searches City (unaffected). SubredditMediaFeed:1200 uses Location::query() DB (leave).
- AppSettings.php: schema-driven ($sections + $schema arrays); `get/set`; comma-sep text lists are the convention (e.g. subreddit_similarity_excluded_usernames). booted() forgets cache on updated. Ad-hoc (non-schema) keys are preserved by getAll/checkForInitKeys.
- Observers registered in providers: e.g. ScraperV2ServiceProvider:72 `Favorite::observe(...)`. Register AppSettings observer in AppServiceProvider::boot().
- console.php: no APP_ENV gate; scheduler runs on web host; use `->onOneServer()->withoutOverlapping()->runInBackground()`. Toggles read at top of file. Schedule::command(Cls::class, ['--flag']) passes args.
- Clone DB state: migrate up-to-date except 1 unrelated pending; **locations=46,949, location_reddit_user=197,765**, is_search_visible col not present. Tests: MySQL `testing` DB, RefreshTestDatabase pattern; command-test exemplars BackfillLocationRedditUserPivotTest, AccountNetworkDiagnosticsCommandTest.

## DESIGN DECISION (Scout reconciliation) — OPTION (b), filterable-attribute + public-only filter
Rationale: gating shouldBeSearchable on the flag (option a) would drop junk from the Meili index, breaking AdminLocationSearchPanel's merge tool (admins search junk to merge it — the future cleanup cycle spec says to preserve). So: keep ALL non-merged locations in the index; add `is_search_visible` as a filterable attribute; filter `->where('is_search_visible', true)` ONLY at public call sites. Matches existing `->where('top_level', true)` convention exactly. Sweep still re-indexes flipped rows so the Meili attribute updates.

## Files to CREATE
1. **Migration** `..._add_is_search_visible_to_locations_table.php`: `boolean('is_search_visible')->default(false)->index()->after('slug')` (instant + online index; 46K/206K rows trivial).
2. **Service** `App\Services\Locations\LocationQualification`: `conditionKeys()` (the 6 tunable keys), `conditions()` (reads AppSettings → min_name_length,min_authors,min_posts,blocklist[],allowlist[],require_city_id), and `qualifies($locString,$merged,$cityId,$authors,$posts,$cond): bool` predicate. Predicate: merged→false; lower(trim(loc_string)) in blocklist→false; in allowlist→true; else strlen>=min_name_length && authors>=min_authors && posts>=min_posts && (!require_city_id || city_id!==null).
3. **Command** `App\Console\Commands\Locations\RecomputeLocationQualification` sig `lounge:recompute-location-qualification {--force} {--if-requested} {--dry-run} {--reindex-all} {--chunk=5000} {--resume-from=}`. Master gate `location_qualification_enabled` (unless --force). Chunk by id range: aggregate location_reddit_user for chunk → compute newFlag per row in PHP → collect flips vs current → bulk `DB::table('locations')->whereIn(id,...)->update([is_search_visible=>0/1])` (no model events) → Scout reconcile the flipped **non-merged** ids: `Location::whereIn('id',$ids)->get()->searchable()` guarded by ScoutIndexing::enabled(). `--reindex-all` re-indexes all non-merged (bring-up). `--if-requested`: run only if AppSettings `location_qualify_rebuild_requested_at` set, then forget it. Stamp `location_qualify_last_swept_at`. Report scanned/qualified/flipped/reconciled. Idempotent (no-op run = 0 Scout work), resumable via --resume-from.
4. **Observer** `App\Observers\AppSettingsQualificationObserver` `updated(AppSettings $s)`: if `$s->key ∈ LocationQualification::conditionKeys()` && `$s->wasChanged('value')` → `AppSettings::set('location_qualify_rebuild_requested_at', now ts)`. No loop (bookkeeping key not a condition key). Register in AppServiceProvider::boot() via `AppSettings::observe(...)`.

## Files to EDIT
- `app/Models/Location.php`: add `is_search_visible => (bool)$this->is_search_visible` to toSearchableArray(); add `'is_search_visible'` bool cast. (Leave shouldBeSearchable AS-IS.)
- `config/scout.php`: add `'is_search_visible'` to Location filterableAttributes (line ~313).
- `app/Models/AppSettings.php`: add `locations` section to $sections; add schema keys: `location_qualification_enabled`(bool,true), `location_qualify_min_name_length`(int,3), `location_qualify_min_authors`(int,5 — evidence: floor 5 keeps ~21.8K), `location_qualify_min_posts`(int,0=disabled), `location_qualify_blocklist`(text,'everywhere,mf'), `location_qualify_allowlist`(text,''), `location_qualify_require_city_id`(bool,false). (Bookkeeping keys rebuild_requested_at/last_swept_at are ad-hoc, NOT in schema.)
- `app/Providers/AppServiceProvider.php`: register observer in boot().
- `routes/console.php`: read `location_qualification_enabled` at top; schedule daily `->dailyAt('04:15')->withoutOverlapping()->onOneServer()->runInBackground()` + every-5-min `['--if-requested']` same guards.
- Run `composer ide-helper` after migration for docblocks.

## Tests to ADD (tests/Feature)
- LocationQualification predicate unit + RecomputeLocationQualificationTest (seeds locations + location_reddit_user, asserts flags flip correctly, blocklist/allowlist/length/author floors, merged→false, idempotency, --if-requested gating). AppSettings observer requests rebuild on condition-key change.

## Low-hanging-fruit / characterization
Evidence in spec already gives the tuned default: distinct-author floor 5 keeps ~21.8K of 206K. Will attempt read-only prod characterization (bin/diag / SSH PHP pattern) to record distribution + confirm; won't block build. Defaults accepted by operator (event #3206): flag `is_search_visible`, daily cadence, min_name_length 3, min authors/posts tunable.

## Verify + wrap
`php artisan test` (new suites) + `./vendor/bin/pint`. Commit(s) on feature/authors-optimization each with `Brief: #242` trailer, NO Co-Authored-By. On done: brief_append summary+SHAs then brief_dispatch_complete dispatch_request_id=127.

Acknowledging operator PAUSE for Solo restart — going idle now; will resume implementation after restart.
[17] dispatched lounge-orchestrator: Dispatch request #127 marked cancelled.

Recommended linked context:
{
    "todos": [],
    "scratchpads": []
}

Execution notes:
- Treat the brief as the source of truth.
- Keep work scoped to this dispatch request.
- Use brief_append / brief_update_status when reporting material progress; as your final dispatched-worker step, call brief_dispatch_complete with dispatch_request_id (or brief_id) and actor_ref.
- Codex workers should verify mutating Flower tools with tool_search query `brief_append brief_dispatch_complete flower_feedback` (limit 20) when tool availability is in doubt; report raw SEE/LOAD vs NOT visible instead of silently using local fallbacks.
- Add a git commit trailer `Brief: #242` to every commit for this brief so flower can exact-link commits back to the brief.
- Need an operator call while working this brief? A question ABOUT THIS BRIEF -> brief_ask(242, ...); a standalone decision not tied to the brief -> decision_ask(...). Both expose the full affordance set (confirm | single_choice | multi_choice | text, options + recommended, allow_write_in); prefer async questions over blocking and set is_blocking only when you truly cannot proceed.

You are being dispatched from flower Brief #242: Location qualification foundation + low-hanging-fruit cleanup (prereq for /authors optimization)

Recall pointer:
- Use recall_brief with id 242 for the full folder if you need provenance.

Target:
- project: lounge (/Users/mikeferrara/Documents/code/lounge)
- branch: feature/authors-optimization
- worktree: not specified
- kind: fresh

Current brief spec:
# Location qualification foundation + low-hanging-fruit cleanup

## Goal
Establish a **precomputed, config-driven "qualified location" set** as the clean foundation the /authors optimization epic (#239) builds on. Replace doc-30's maintained live query-time counts with a boolean flag on `locations`, driven by tunable AppSetting conditions, kept fresh by a scheduled sweep, and reconciled into Meilisearch via Scout — so location search/typeahead query a qualified set, not the raw ~206K rows (~94% junk). Non-destructive: junk is flagged out of discovery, not deleted. **Sequenced FIRST; #239 depends on it.**

## Background / evidence
- `locations` ≈ 206K rows on prod, ~94% junk (`everywhere`, `mf`, one-off typos) per the /authors-optimization design (doc 30 / 00-PRODUCTION-FACTS). A distinct-author floor of 5 would keep ~21.8K.
- Today there is no clean set — typeahead/search surface the raw table. Building proximity/search/favorites on this means every downstream piece inherits the junk.
- Operator principle: clean the foundation before building the layer; a solid config-driven foundation, not an exhaustive cleanup.

## Design — qualified flag + config + sweep + Scout
1. **Flag column** on `locations` — a boolean (name TBD at build, e.g. `is_search_visible`; avoid collision with the author `only_indexed` concept). Indexed. Defaults false; the sweep sets it.
2. **AppSetting-driven qualification conditions** (tunable, not hardcoded):
   - `location_qualify_min_name_length` (default 3 — drops `mf`, keeps `PDX`; allowlist any legit 2-char names if found).
   - `location_qualify_min_posts` and/or `location_qualify_min_authors` (low threshold; final defaults tuned by the low-hanging-fruit pass).
   - `location_qualify_blocklist` (seed with known junk: `everywhere`, `mf`, …; extensible).
   - Optional `city_id IS NOT NULL` / coordinate-presence gate if useful for proximity.
3. **Scheduled sweep** — `lounge:recompute-location-qualification` (or similar), scheduled (default daily, off-peak; cadence configurable). Recomputes the flag for all locations against current conditions. Chunked, idempotent, resumable, on the maintenance worker via **inline scheduler — NOT a queued job** (queue-reference 15s-timeout gotcha).
4. **Rebuild-on-config-change** — an AppSetting observer/event: when any `location_qualify_*` key changes, trigger a re-sweep (or mark dirty for the next run) so the qualified set tracks config.
5. **Scout → Meilisearch reconciliation** — location search/typeahead query the qualified set. Either (a) `Location` uses Scout with `shouldBeSearchable()` gated on the flag (unqualified locations drop out of the Meili index), or (b) a flag filter in the search query — pick whichever matches existing Scout usage here. Reconcile the Meili index off the flag like other models.

## Low-hanging-fruit pass (part of this brief)
- Characterize the junk on prod (read-only): distribution by name length, by post/author count, most-common junk names, obvious non-place strings — via `bin/diag` / the read-only SSH PHP-script pattern.
- Tune the initial condition defaults from that characterization.
- Investigate whether tooling helps identify junk at scale (typo / non-place detection) — optional; note findings, don't over-build.
- **Non-destructive:** sets flags/conditions only; does NOT delete or merge location records (deeper cleanup/merge is a separate future cycle).

## Constraints
- **Discovery only.** The flag gates search/typeahead visibility; `resolveFromSlug` reachability stays untouched (existing links keep working).
- Non-destructive (flag, don't delete/merge).
- Schema add is INSTANT (one boolean col + index, online/off-peak on the memory-tight prod box).
- Sweep/reconcile run inline-scheduled, not queued.

## Acceptance criteria
- `locations` has the qualified flag + index; the sweep populates it from AppSetting conditions.
- Changing a `location_qualify_*` AppSetting triggers/permits a rebuild and the set updates.
- Location search/typeahead return only qualified locations; Meili index reconciled off the flag.
- A junk-characterization summary + tuned default thresholds are recorded.
- Zero deletions/merges of location records; reachability unaffected.

## Relationship
Prereq for #239 (/authors optimization); replaces #239's former C3 minimal floor. Broader app-wide location-search centralization is a separate later round.

## Open detail questions (low-stakes; defaults proposed, tunable post-ship)
- Flag column name (`is_search_visible` vs `is_qualified`) — default `is_search_visible`, final at build.
- Sweep cadence — default daily off-peak, configurable.
- Initial thresholds — min name length 3; min posts/authors tuned from the characterization pass.

Recent/key trace events:
[1] participant_joined lounge-refine: (no body)
[2] note_added lounge-refine: **Operator ask (2026-07-04, from #239 scope discussion):** Before building the /authors optimization layer (proximity/search/favorites) on top of a junk-filled `locations` table (~206K rows, ~94% junk per the authors-optimization design — `everywhere`, `mf`, one-off typos), do a foundational qualification + low-hanging-fruit cleanup pass FIRST, so everything downstream searches a clean set. Operator: "cleaning this up is likely something we should approach before we start building a whole huge layer on top of it."

**Proposed approach (operator's, refine concurs — this is the better design than doc-30's live query-time counts):**
- A **qualified boolean flag** on `locations` (name TBD — avoid collision with the existing `only_indexed` author concept; e.g. `is_search_visible` / `is_qualified`).
- **AppSetting-driven qualification conditions** (min name length — e.g. ≥3 to drop `mf` but keep `PDX`; min post/author count; a blocklist of known-junk names like `everywhere`). Tunable config, not hardcoded.
- A **scheduled sweep** that (re)computes the flag against the conditions (daily or configurable) + an **AppSetting observer/event** that rebuilds the set when the conditions change.
- **Scout → Meilisearch reconciliation off the flag** (same pattern as other models here), so location search/typeahead queries the qualified set directly.
- A **low-hanging-fruit identification pass**: characterize the worst offenders / easiest-to-filter values (and investigate whether any tooling helps), to tune the initial condition set. Not an exhaustive cleanup — a solid, config-driven foundation.

**Architecture note:** recommend the boolean-flag + Scout approach over a separate "working-set table" — the flag + Scout filter achieves the same clean set without a parallel table to join/maintain, and matches the codebase's existing Scout/Meili reconciliation pattern.

**Sequencing:** this REPLACES the /authors epic's minimal query-time floor (former C3) and is done FIRST. The /authors optimization epic (#239) depends on this qualified set (proximity + typeahead search it). The broader app-wide location-search centralization is a SEPARATE later round (operator: audit the messy search landscape and unify selectively, after /authors).
[3] status_change lounge-refine: (no body)
[4] plan_proposed lounge-refine: # Location qualification foundation + low-hanging-fruit cleanup

## Goal
Establish a **precomputed, config-driven "qualified location" set** as the clean foundation the /authors optimization epic (#239) builds on. Replace doc-30's maintained live query-time counts with a boolean flag on `locations`, driven by tunable AppSetting conditions, kept fresh by a scheduled sweep, and reconciled into Meilisearch via Scout — so location search/typeahead query a qualified set, not the raw ~206K rows (~94% junk). Non-destructive: junk is flagged out of discovery, not deleted. **Sequenced FIRST; #239 depends on it.**

## Background / evidence
- `locations` ≈ 206K rows on prod, ~94% junk (`everywhere`, `mf`, one-off typos) per the /authors-optimization design (doc 30 / 00-PRODUCTION-FACTS). A distinct-author floor of 5 would keep ~21.8K.
- Today there is no clean set — typeahead/search surface the raw table. Building proximity/search/favorites on this means every downstream piece inherits the junk.
- Operator principle: clean the foundation before building the layer; a solid config-driven foundation, not an exhaustive cleanup.

## Design — qualified flag + config + sweep + Scout
1. **Flag column** on `locations` — a boolean (name TBD at build, e.g. `is_search_visible`; avoid collision with the author `only_indexed` concept). Indexed. Defaults false; the sweep sets it.
2. **AppSetting-driven qualification conditions** (tunable, not hardcoded):
   - `location_qualify_min_name_length` (default 3 — drops `mf`, keeps `PDX`; allowlist any legit 2-char names if found).
   - `location_qualify_min_posts` and/or `location_qualify_min_authors` (low threshold; final defaults tuned by the low-hanging-fruit pass).
   - `location_qualify_blocklist` (seed with known junk: `everywhere`, `mf`, …; extensible).
   - Optional `city_id IS NOT NULL` / coordinate-presence gate if useful for proximity.
3. **Scheduled sweep** — `lounge:recompute-location-qualification` (or similar), scheduled (default daily, off-peak; cadence configurable). Recomputes the flag for all locations against current conditions. Chunked, idempotent, resumable, on the maintenance worker via **inline scheduler — NOT a queued job** (queue-reference 15s-timeout gotcha).
4. **Rebuild-on-config-change** — an AppSetting observer/event: when any `location_qualify_*` key changes, trigger a re-sweep (or mark dirty for the next run) so the qualified set tracks config.
5. **Scout → Meilisearch reconciliation** — location search/typeahead query the qualified set. Either (a) `Location` uses Scout with `shouldBeSearchable()` gated on the flag (unqualified locations drop out of the Meili index), or (b) a flag filter in the search query — pick whichever matches existing Scout usage here. Reconcile the Meili index off the flag like other models.

## Low-hanging-fruit pass (part of this brief)
- Characterize the junk on prod (read-only): distribution by name length, by post/author count, most-common junk names, obvious non-place strings — via `bin/diag` / the read-only SSH PHP-script pattern.
- Tune the initial condition defaults from that characterization.
- Investigate whether tooling helps identify junk at scale (typo / non-place detection) — optional; note findings, don't over-build.
- **Non-destructive:** sets flags/conditions only; does NOT delete or merge location records (deeper cleanup/merge is a separate future cycle).

## Constraints
- **Discovery only.** The flag gates search/typeahead visibility; `resolveFromSlug` reachability stays untouched (existing links keep working).
- Non-destructive (flag, don't delete/merge).
- Schema add is INSTANT (one boolean col + index, online/off-peak on the memory-tight prod box).
- Sweep/reconcile run inline-scheduled, not queued.

## Acceptance criteria
- `locations` has the qualified flag + index; the sweep populates it from AppSetting conditions.
- Changing a `location_qualify_*` AppSetting triggers/permits a rebuild and the set updates.
- Location search/typeahead return only qualified locations; Meili index reconciled off the flag.
- A junk-characterization summary + tuned default thresholds are recorded.
- Zero deletions/merges of location records; reachability unaffected.

## Relationship
Prereq for #239 (/authors optimization); replaces #239's former C3 minimal floor. Broader app-wide location-search centralization is a separate later round.

## Open detail questions (low-stakes; defaults proposed, tunable post-ship)
- Flag column name (`is_search_visible` vs `is_qualified`) — default `is_search_visible`, final at build.
- Sweep cadence — default daily off-peak, configurable.
- Initial thresholds — min name length 3; min posts/authors tuned from the characterization pass.
[5] note_added lounge-refine: Operator accepted the detail defaults (2026-07-05): flag name `is_search_visible`, daily sweep cadence, min name length 3, min post/author count TBD-from-characterization — "we'll likely address and adjust during the build." So the "Open detail questions" are resolved as accepted defaults, tunable at build time. #242 spec is settled/dispatch-ready.
[6] status_change lounge-refine: (no body)

Recommended linked context:
{
    "todos": [],
    "scratchpads": []
}

Execution notes:
- Treat the brief as the source of truth.
- Keep work scoped to this dispatch request.
- Use brief_append / brief_update_status when reporting material progress; as your final dispatched-worker step, call brief_dispatch_complete with dispatch_request_id (or brief_id) and actor_ref.
- Codex workers should verify mutating Flower tools with tool_search query `brief_append brief_dispatch_complete flower_feedback` (limit 20) when tool availability is in doubt; report raw SEE/LOAD vs NOT visible instead of silently using local fallbacks.
- Add a git commit trailer `Brief: #242` to every commit for this brief so flower can exact-link commits back to the brief.