diff --git a/phase-ends/current/logs/T5.c5.md b/phase-ends/current/logs/T5.c5.md
new file mode 100644
index 0000000000..299f2754a7
--- /dev/null
+++ b/phase-ends/current/logs/T5.c5.md
@@ -0,0 +1,59 @@
+# T5.c5: I/M sweep completion, residue, gate
+
+## Waiting on c3 cycles
+- c3 cycle `restruct_cycle.sh 2 20 200 S ov_*` (pid 2091853) ran to exit; drawable ran out at S13. All batches committed green.
+- The orphaned c3 agent then started `15 17 200 S md_*` (only S15 committed) and `18 19 200 S main` (S18). It then committed its log in bd1428af31. I waited each out and only went ahead after 3 min with no cycle running.
+- Labels S14, S16, S17, S19: nothing drawable, so no batch was committed.
+
+## Batches (drawn / residual by cause; all final N/N identical; R22 check-all 218 passed 0 failed unless noted)
+- S1 (pre-c5, 17 files): drawn 14 / resid 279
+- S2 6908e68e41: 221 / 8923 {COMPILE-ERROR 269, DIFFERS 828, NO-SCALAR-EXTERN 2548, REKEYS 108, WIDTH 5166, R3-COMPILE-ERROR 1, R3-NOT-CONTIGUOUS 3}
+- S3 e9a254d8b0: 511 / 3579 {CE 127, DIFFERS 1387, NSE 453, REKEYS 290, WIDTH 1318, R3-NC 4}
+- S4 cd92936bd8: 62 / 2648 {CE 175, DIFFERS 1189, NSE 342, REKEYS 232, WIDTH 710}
+- S5 6a6d776ea3: 137 / 2419 {CE 351, DIFFERS 70, NSE 485, REKEYS 478, WIDTH 1032, R3-CE 3}
+- S6 902229d4b4: 127 / 1592 {CE 4, DIFFERS 120, NSE 395, REKEYS 168, WIDTH 905}
+- S7 7ff1e7b4ca: 142 / 1310 {CE 136, DIFFERS 74, NSE 751, REKEYS 113, WIDTH 235, R3-NC 1}
+- S8 589ceabeb3: 134 / 1153 {CE 97, DIFFERS 137, NSE 770, REKEYS 43, WIDTH 105, R3-NC 1}
+- S9 4d680e6874: 47 / 936 {CE 3, DIFFERS 104, NSE 535, REKEYS 139, WIDTH 154, R3-NC 1}
+- S10 5377c95ad3: 121 / 652 {CE 3, DIFFERS 13, NSE 414, REKEYS 87, WIDTH 132, R3-NC 3}
+- S11 0816ddaf30: 63 / 5429 {CE 3, DIFFERS 11, NSE 5294, REKEYS 11, WIDTH 110}
+- S12 299b1c1b4e: 64 / 5527 {CE 3, DIFFERS 10, NSE 5393, REKEYS 10, WIDTH 111}
+- S13 05b01fb549 (56 files, ov drained): 19 / 1315 {DIFFERS 1, NSE 1280, WIDTH 33, R3-NC 1}
+- S15 8a09f83c78 (md_*, 46 files): 9 / 147 {CE 2, DIFFERS 10, HEADER-EXTERN 3, NSE 73, REKEYS 46, WIDTH 13}
+- S18 b06a6794e5 (main, 6 files): 0 / 38 {DIFFERS 3, NSE 6, REKEYS 18, WIDTH 11}; no src change, so the previous R22 result stands
+- c5 shared headers (HEADERS=1, `--only shared`):
+  - S20 cf1c2d6d44 (pilot, 5 headers): 0 / 149 {HEADER-EXTERN 131, NSE 9, WIDTH 9}; R22 stands
+  - S21 4da554f70: 2 / 255 {CE 28, DIFFERS 2, HDR-EXT 162, NSE 38, WIDTH 25}
+  - S22 ceb263338: 1 / 71 {CE 15, HDR-EXT 32, NSE 9, REKEYS 1, WIDTH 14}
+  - S23 283376617: 2 / 49 {CE 19, HDR-EXT 8, NSE 10, WIDTH 12}
+  - S24: nothing drawable.
+- Red or uncommitted batches: none. `--restore` was not needed.
+
+## Deviation: shared headers needed a ladder pass
+- The brief routed src/shared to `--residue` only. My first residue pass wrote RESIDUAL(GADDR) 464 and RESIDUAL(GLOBAL) 65 for shared.
+- Those are global-base sites the ladder never drew. `plan_S_im` draws `.h` TUs only with `--headers` (retriever: tools/restruct.py:3775). Without `--headers`, `--plan --only shared` reports 0 drawable.
+- Fix:
+  - The residue rows were uncommitted, purely additive and mine only, so I reverted the ledger with `git checkout`.
+  - `--plan --headers --only shared` then showed 176 drawable headers with 376 sites.
+  - I ran the header cycles S20 to S23. They were fast (mostly HEADER-EXTERN refusals, few compiles).
+  - Then I ran the residue again.
+- One GADDR residual remains: src/800_b_2.c:4639, `gaddr:D_800A4F17`, `<file-scope>` (fn null). The ladder never draws file-scope sites (plan_S_im needs fn), so it is legitimately non-drawable.
+- `--plan --rung S --forms I,M` gives 0 drawable for ov_*, md_*, main and shared, both with and without `--headers`.
+
+## Residue (final, `--rung S --residue --forms I,M --only <seg> --label t5_res_<seg>`)
+- ov_*: 6376 rows, 24000 sites {NO-EDIT 665, NO-TYPE-IN-MAP 773, TYPE-NOT-CANONICAL 22562}
+- md_*: 67 rows, 174 sites {NO-TYPE-IN-MAP 24, TNC 150}
+- main (matches src/resident too): 99 rows, 284 sites {FILE-SCOPE 2, GADDR 1 (file-scope, above), NO-TYPE-IN-MAP 28, TNC 253}
+- shared: 255 rows, 875 sites {NO-TYPE-IN-MAP 21, TNC 689, TYPE-NOT-VISIBLE 165}
+- resident: the earlier check showed 7 uncovered. They are covered by the main rows (`--residue --only resident`: 0 rows).
+
+## Census
+- Before c5 residue (after c3 sweep): `check-casts: raw P=257464 I=49836 X=13790 M=15128 (uncovered P=0 I=12157 X=0 M=13704) ... lying=229 (ledgered 0)`
+- After: `check-casts: raw P=257464 I=49831 X=13790 M=15128 (uncovered P=0 I=0 X=0 M=0); macros CAST_ALIAS=1173 CAST_SIGN=8429 CAST_WIDTH=10410 CAST_MISALIGNED=624 CAST_NONSTRUCT=327 (unbacked 0); lying=229 (ledgered 0)`. rc 1 comes from lying=229 (owned by T6).
+
+## Gate
+- `bash tools/run.sh t5c5_gate -- bash -c 'make clean && make extract-all JOBS=16 && make check-all JOBS=16'`: `check-all: 218 passed, 0 failed of 218`
+
+## Notes
+- harness: `tools/commit_task.sh` refuses the ignored-dir path `.run/P37/restruct/ledger.jsonl` (`git add` without -f). I staged it with `git add -f` and then committed through the script with the log path.
+- workflow: the residue pass covers global-base sites it was never given, and then hides them from `--plan`. Run header cycles (HEADERS=1) before `--residue` for any segment that has `.h` TUs.