feat(remote): seed the drift baseline from the stored snapshot on connect

[veksen]

Closes Query-Doctor/Site#3670.

Goal

Make a stale Production Statistics snapshot heal itself, with no user action.

#195 added drift detection, but it measures against the last dump this process pushed. An analyzer that has never pushed has no baseline and never drifts. That is every project whose snapshot was seeded by hand.

Query Doctor's own project is the case: an analyzer connected for 55 days, heartbeating, that has pushed statistics zero times. Its snapshot was taken by a manual psql dump on 2026-06-18 and six tables added since are uncovered.

What

Before, drift armed only after this process dumped from source at least once.

After, the analyzer adopts the server's stored snapshot as its baseline on connect. The first schema poll then compares the live schema against it and re-dumps if it has fallen behind.

How

Remote.seedStatsBaseline takes a snapshot and records it as the drift baseline. ApiConnection.connectWithReconnect calls getProductionStats() after connecting and passes the result in. That RPC already exists and is already called on the CI path, so there is no contract change.

Two guards, both tested:

• Seeding never pushes. These numbers came from the server. Emitting them would republish a stale snapshot as though it were a fresh dump.
• Seeding is ignored once a baseline exists. Otherwise a reconnect would discard a fresher local baseline for the server's older copy, and re-dump on every reconnect.

The fetch is fire-and-forget. If it fails, drift stays inert until a local dump sets a baseline, which is the behaviour before this change.

Tests

seed-stats-baseline.test.ts, four cases: a seeded baseline makes a new table read as Shape Drift; an empty snapshot leaves drift inert; a second seed does not overwrite the first; seeding emits no statsApplied. No container needed, since seeding only writes to memory.

[github-actions]

Query Doctor Analysis

3 queries analyzed

0 regressed · 0 improved · 0 new · 0 removed

2 pre-existing issues

• SELECT "guests"."id", "guests"."session_id", "guests"."username", "guests"."avatar_path", "guests"."color", "guests"."side", "guests"."audio_recording_path", "guests"."audio_recording_public", "gue...
  index assets(event_id, inserted_at desc)
  cost 31,003,449 → 1,498 (100% reduction)
• SELECT * FROM guest_ip_addresses WHERE ip_address = '127.0.0.1';
  index guest_ip_addresses(ip_address)
  cost 154,402 → 8 (100% reduction)

_{Using assumed statistics (10000000 rows/table). For better results, sync production stats.}

_{More detail → get_ci_run({ runId: "019fa54e-fce8-76a2-bbd6-ca3c06df5b0c" }) · view run · docs}