fix: derive isComplete from isRealtime+pct instead of ponder_sync_is_complete (COW-1008)

[lgahdl]

Summary

• ponder_sync_is_complete never reaches 1 for chains with live block handlers (C1–C5 run indefinitely after historical sync ends)
• isComplete is now computed locally as isRealtime && pct >= 100, which correctly reflects that indexing is fully caught up
• Added regression test: verifies isComplete=true when ponder_sync_is_complete=0 but isRealtime=1 and all blocks processed

Test plan

• pnpm test tests/api/sync-progress.test.ts — 8 tests, all pass
• pnpm typecheck — clean
• Hit /api/sync-progress on staging after deploy; confirm isComplete flips to true once isRealtime=true and historicalBlocksFetchedPct=100

🤖 Generated with Claude Code

[linear-code]

COW-1008

[socket-security]

Review the following changes in direct dependencies. Learn more about Socket for GitHub.

Diff	Package	Supply Chain Security	Vulnerability	Quality	Maintenance	License
	tsx@​4.22.3

View full report

[lgahdl]

Review: COW-1008 — Fix isComplete

Code Quality: PASS — One-line fix with an inline comment explaining the root cause. The comment makes the workaround self-documenting for future maintainers.

Coherence: PASS — Directly addresses the bug: ponder_sync_is_complete never reaches 1 because C1-C5 live block handlers run indefinitely. The local derivation (isRealtime && pct >= 100) is the correct signal for 'all historical blocks processed and now receiving live blocks'.

Functionality: PASS — The test suite now covers the regression: a chain where ponder_sync_is_complete=0 but isRealtime=1 and pct=100 correctly returns isComplete=true. The existing case where isRealtime=1 but pct<100 correctly returns isComplete=false.

[jeffersonBastos]

Superseded by #102 (final-qa → develop). Verified that this PR's changes (COW-1008) are fully contained in the consolidated final-qa branch. Closing to merge everything through the single QA branch per the agreed plan — no work is lost.