fix(remote): publish production statistics from the sync that dumped them

[veksen]

Goal

An analyzer running against a database publishes that database's statistics on its own. Today it never does, so every project costs its queries against the 10M-row assumption for its whole life.

What

refreshStatsIfStale is the only code path that pushes statistics to the server. It returns immediately when statsBaseline is unset. The only thing that set statsBaseline was a push, and the only other way in, seedStatsBaseline, needs a snapshot the server does not have yet. A project with no snapshot never dumps, so it never gets one.

Found on veksen-enterprises/d2armory, whose analyzer has been connected since 2026-07-20. Its latest CI run reports statisticsMode: fromAssumption and statisticsPayloadId: null, and the container logs Statistics refresh skipped: no drift baseline, so nothing can trigger a dump.

How

syncFrom already dumps the source statistics and hands them to the optimizer. It now also passes them to recordSourceStatistics, extracted from applyStatistics, which sets the baseline and emits statsApplied. The first sync at boot publishes, and drift keeps the snapshot current after that.

decideStatsStrategy used to skip the dump below STATS_ROWS_THRESHOLD. It now dumps either way and returns the dump alongside the mode. The threshold still decides what the optimizer costs against; it no longer decides whether production's numbers reach the server. StatsResult.dump carries only statistics measured from the source, so a static mode loaded from a file still publishes nothing.

Republishing on every boot costs nothing on the server: setFromAnalyzer hashes the tables and a match only moves confirmed_at.

Tests

src/remote/publish-stats-on-sync.test.ts syncs from a container holding 200 rows, which is under the threshold, and asserts statsApplied carries reltuples: 200. It fails on main with zero emissions.

Full suite: 455 tests across 44 files, all passing. tsc --noEmit clean.

[github-actions]

Query Doctor — 6 successful checks

 Cost regression — No query went up more than 5%
 Untested data access — No changed data-access file without a test
 New query — No new queries
 New query with index recommendation — No new query ships an index recommendation
 Schema drift — No schema changes
 High-value nudge — No index or rewrite past the threshold

---

_{More details via MCP → get_ci_run({ runId: "019ffe21-35b9-757c-a149-05d0a71fdc77" }) · view run · docs
3 queries read against main on assumed statistics of 10,000,000 rows per table. Sync production stats for costs measured against your real data.}