Was: malformed event (bad JSON / truncated / corrupted bytes) wedged the
node's checkpoint forever — every cycle re-tried, logged, never advanced
past the bad file; all subsequent good events for that node lost.
Now: first parse failure -> atomic os.replace to STATE_DIR/observer_failed_events/<node>/
with collision handling. Checkpoint advances, downstream events flow.
Move failures themselves are logged but don't crash the loop.
Complementary to yesterday's atomic_write_json fix (state files);
this addresses the same race-pattern on event files instead.
Regression test asserts: bad event quarantined to failed_events dir,
removed from hot path, subsequent good event processed (node online),
checkpoint moves to good event.