fix(operator-trend): use any() for rererestore freshness signal flag (T3-2 phase 5b)

[saagpatel]

The drift

closure_forecast_reset_reentry_rebuild_reentry_restore_rererestore_freshness_for_target computed its has_fresh_aligned_recent_evidence flag differently from every sibling tier:

# rererestore (outlier)                              # all 6 siblings
freshness_status == "fresh"                          any(
and aligned_recent_event_count >= 2                      side(event) == current_side ...
                                                         for event in relevant_events[:2])

Every other freshness builder (reset_reentry, reset_reentry_rebuild, reacquisition, rebuild_reentry, rebuild_reentry_restore, rebuild_reentry_restore_rerestore) uses an any(...) existential over the first two recent events. rererestore was the lone outlier — a stricter count-threshold + global-freshness gate.

The flag is load-bearing: read 49 times in operator_resolution_trend.py, gating churn_status and persistence decisions.

The fix

Convert the flag to the sibling any(...) form over rererestore's own aligned predicate; drop the now-dead aligned_recent_event_count sum. The flag now fires on any recent aligned event like its siblings (previously: both of the first two and overall freshness "fresh").

Caught by the phase-4 net

The composer golden (#79) caught it: a 3-line diff (has_fresh_aligned_recent_evidence: false → true in 3 scenarios). No existing test pinned the old value — the path was uncharacterized, which is why the drift survived.

Verification

• Full suite green (2539 passed, 2 skipped) — zero mirror-test breakage
• mypy on edited source clean; ruff check src/ tests/ clean
• Golden diff is the minimal 3-line behavior change

Note for collapse planning

This removes the clear aggregation outlier. A residual predicate difference remains across the freshness tiers (rebuild keys on signal_label != "neutral", rerestore on per-event freshness == "fresh"), so a full freshness-family collapse will still need a predicate-unification decision — tracked separately, not in scope here.