Resolve writeSyncAnchor signature conflict: PR garrytan#1430's pullFailed param and
upstream garrytan#1623's newestContentEpochMs param both added a 5th positional arg to
the same function. Merged to take both — newest_content_at (git-intrinsic HEAD
committer time) stamps regardless of pull outcome; last_sync_at (the
"observed upstream" freshness signal) stays gated on pullFailed. All 3 call
sites pass both args. tsc --noEmit clean.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

garrytan#1794 (garrytan#1430)

Upstream garrytan#1794 (resumable sync) covers the no-op-freshness half of garrytan#1430 but
dropped our pullFailed gate, so last_sync_at stamps even when the git pull
failed (doctor reads a never-observed source as fresh). Re-threads
pullAttemptedAndFailed through writeSyncAnchor + performFullSync. Verified:
fixes the pull-failure suppression cases in
test/sync-no-op-touches-last-sync-at.test.ts (4/5 pass; the 1 residual is a
pre-existing test-vs-garrytan#1794 mismatch on the plain no-op path, fails on the
pure-upstream base too).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ock (#dcb38654) & pull-failure suppression (garrytan#1430) into local/integration

Resolves the stuck auto-ingest fork-sync: conflicts in db.ts/postgres-engine.ts/sync.ts
resolved to upstream's now-canonical fixes (f3ade6c ownership-token, ec5fed2 reconnect,
garrytan#1794 resumable-sync); two surviving local fixes upstream lacks re-grafted on top.
tsc clean; connection-resilience 25/25; garrytan#1430 suppression tests pass.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…_sync_at advance

The fork-sync re-graft initially carried only the pull-failure suppression half
of garrytan#1430 (mistakenly assuming upstream garrytan#1794 absorbed the no-op-advance half; it
did not — the up_to_date early-return still skips the freshness write). Adds the
missing no-op advance so doctor's sync_freshness stops misreading quiet sources
as stale. Matches the refreshed PR garrytan#1430.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

Resolves the fork-sync conflict that the self-heal couldn't auto-resolve
(typecheck gate failed on keep-both markers in src/core/search/mode.ts).

Conflict classes:
- mode.ts / search.ts: additive on both sides — our autocut weak-top floor
  (acmts) + upstream's v0.43 relational-recall arm (rel/reld) now coexist.
  Both sides independently bumped KNOBS_HASH_VERSION 9->10; merged code carries
  BOTH contributions, so bumped to 11 to force a clean one-time cache cold-miss
  (a merged-v10 hash composition matches neither published v10).
- sync.ts: took upstream's garrytan#1794/garrytan#1970 resumable-sync rewrite wholesale, then
  re-grafted our garrytan#1430 pull-failure-freshness fix onto it (writeSyncAnchor
  pullFailed param + no-op-path last_sync_at bump). Verified upstream does NOT
  cover garrytan#1430 before re-grafting (its writeSyncAnchor is 5-arg).
- 4 test files: unioned the mode-bundle expectations + bumped KNOBS_HASH_VERSION
  assertions to 11.

Verification: typecheck clean; search-mode/knobs suites 138/138; all 31
sync test files green in isolation (garrytan#1430 behavior tests pass).

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

…ped from merge fc957c7

The merge staged upstream's sync.ts via 'checkout --theirs' but the garrytan#1430
re-graft (writeSyncAnchor pullFailed param + no-op-path last_sync_at bump +
pullAttemptedAndFailed threading) was applied to the working tree only and
never re-staged, so fc957c7 captured upstream's version WITHOUT garrytan#1430 — the
exact regression the merge resolution was meant to avoid. Tests passed because
bun reads the working tree, not the index. This commits the validated grafted
version.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>