sync-lww-conflict-noise

Answer

Diagnosis complete and verified. Deliverable: ~/SOLUTION.md.

Root cause

plan_sync handled equal mtimes with a per-direction tie-break (local wins on Push, remote wins on Pull) and logged an LWW conflict. But transfers align mtimes (dest = src), so equal mtimes are the normal steady state — not a conflict. The tie-break re-planned every unchanged file on every run forever.

Fix

Skip equal-mtime pairs entirely in plan_sync:

if l.mtime == r.mtime:
    continue   # steady state: NO-OP, not a conflict

Divergent files are untouched behaviorally: newer side wins and logs exactly one LWW row.

Verification

Built a faithful minimal model (sync_lww_repro.py) + assertions (verify_fix.py) and ran them:

• Buggy planner: 3 transfers / 3 conflicts every round, both directions.
• Fixed planner: 0 transfers / 0 conflicts on unchanged sets, both directions.
• Changed file: transfers once, logs one conflict row, then converges to silence.

ALL ASSERTIONS PASSED. The doc includes the exact code, runnable verification, and a regression test.