diff --git a/phase-ends/CURRENT_PHASE.md b/phase-ends/CURRENT_PHASE.md
index 9bbe992..bba0a5d 100644
--- a/phase-ends/CURRENT_PHASE.md
+++ b/phase-ends/CURRENT_PHASE.md
@@ -177,3 +177,28 @@ excluded_no_extent=0             excluded_trapping_arith=50
   construction as gate-green).
 - Negatives index pending: worker C's 0x80010418/0x800101C7C, worker A's
   0x800F4B54/0x80019700 — extract to config/near_match_negatives.tsv at P9-T8.
+
+## Cycle 2 close & P9-T4 checkpoint (crossed)
+
+| Cycle | New bodies | Total bodies | New regions | Total regions | Result |
+|---|---|---|---|---|---|
+| c2 close (all workers rotated) | — | **297** | — | **306** | gate MATCH; P9-T4 checkpoint (≥290) crossed |
+
+- All three workers rotated out cleanly with handoffs: B (13, budget at cycle-2),
+  C (28, budget), A (108, +79 bodies, budget). No replacement session available;
+  the coordinator absorbs partitions B and C (A's remaining partition is still
+  in-tree but unstaffed).
+- Worker A handoff: 34 negatives (25 match-class + 9 bad-extent triage), 8
+  levers, two high-value unattempted GTE rows (0x80101C2C, 0x80102FA4),
+  undiagnosed-length lead 0x800FDE54. Silent duplicate-claim lesson recorded
+  (0x8010A748: worker C handoff + worker A both wrote byte-identical-but-textual-
+  different sources; merge cannot see it — ledger note: a ` M` on a tracked
+  src/ file at merge time means a second session wrote the same address).
+- **Bad-extent classification implemented** (worker A class, coordinator-
+  verified and disjoint from the corpus): excluded_bad_extent_start=8; worklist
+  regenerated to 1,570 eligible (tier 1=277, tier 2=1,290, tier 3=3); named
+  exclusions now 6 (0x80012A10/0x80012AE0 family, 0x8008A198 maspsx-blocker,
+  gp-thunks, 0x8005DEF8).
+- P9-T5's trapping-arithmetic lead (worker B handoff) verified: 275 trapping
+  words in un-extented gaps vs 661 inside extents — the finding-26 census is
+  under-measured; the gap population is a bounded P9-T5 target.