recall_search exact todo query buried exact completed todo behind unrelated todos with flat scores

After syncing/indexing Solo todos, recall_search project=flower sources=[todo] for "PTY feasibility spike can PHP own long-lived interactive PTYs node-pty Go sidecar" returned the exact PTY todo only at rank 13 with score=1, behind unrelated P0/P1/P2 todos also score=1. The hit content/currency is correct once found, but exact keyword relevance is poorly ranked when todo chunks have flat scores/pending placeholder vectors.

{
    "tool": "recall_search",
    "query": "PTY feasibility spike can PHP own long-lived interactive PTYs node-pty Go sidecar",
    "scope": "project",
    "project": "flower",
    "sources": [
        "todo"
    ],
    "expected": "exact title/body keyword match should appear near the top",
    "observed": "exact closed PTY todo present only after limit=50; unrelated todos ranked above it with score=1"
}

created

5d ago

triaged

5d ago

resolved

—

resolved by

flower-ops

resolution
self-healed: reproduced now — exact PTY todo ranks #1 (score 0.9999), scores properly differentiated, not flat. Was a transient pending-embed window (placeholder/zero vectors -> flat scores) before real todo embeddings landed; same family as #6 (placeholder-vector ranking degradation, deferred backlog). No fix routed. If pending-embed ranking degradation recurs, improvement = keyword-only fallback ranking while vectors pending (ties to #6).