Sesh production indexing boundary and incremental sync

A company-wide Sesh deployment should keep Coding Sessions/TimeTracker as the canonical source, treat Sesh as a rebuildable derived FTS+vector index, and expose it through Sherlock/Cosmos as the ACL-constrained access layer. Do a one-time allowlisted backfill, then ingest incrementally by stable session/message IDs, content hashes, a durable cursor, and tombstones; prioritize deletions and policy revocations, reconcile watermarks periodically, and reserve full blue/green rebuilds for parser/redactor/chunker/model/schema version changes. Caveat: the benchmark thread validated filter isolation and mechanics on a frozen 48-session cohort, but semantic Recall@5 remains unmeasured pending 12 memory-authored questions and independent adjudication. Provenance: Cosmos Coding Sessions session 01a05e70-0865-78e3-ab3f-5f3b01390eba, “Plan company-wide Sesh benchmark,” repository e3-solutions/sesh (2026-09-01–02).