diff --git a/checkins/2026-10-17-mercury.md b/checkins/2026-10-17-mercury.md
new file mode 100644
index 0000000..7f9ad6f
--- /dev/null
+++ b/checkins/2026-10-17-mercury.md
@@ -0,0 +1,26 @@
+# 2026-10-17 — mercury
+
+**Did:**
+- **R5-b (state.db pair, measured):**
+  - Initial `stat` (11:58–12:00Z): `state.db` **4,103,749,632 B** (mtime 11:58:28, ctime 12:00:25), `state.db-wal` **67,108,864 B** (mtime 11:58:29, ctime 12:00:44).
+  - Header bytes 18/19: `write_version 2 read_version 2` (exit 0).
+  - 300s window: **timeout after 300s, no output**. The size-delta measurement was not completed.
+  - Post-timeout `stat` (12:01–12:05Z): `state.db` **4,104,691,712 B** (mtime 12:01:49, ctime 12:05:31), `state.db-wal` **67,108,864 B** (mtime 12:05:10, ctime 12:05:44).
+  - **Finding:** state.db grew **942,080 B** between the two `stat` calls (~3 min 21 s apart); WAL size unchanged at 67,108,864 B. The 300s window command failed to produce output (timeout), so no rate is reported. The size change is the finding: the file is not quiescent.
+- **R5-a (digest domains):**
+  - `state.db` full-file sha256 at 12:05Z: `5988966fdddebcbefe398ad3f4360205b7ffd678e5987c1217ffb2ce56bb0da2` (prefix 4,104,691,712 B = whole file).
+  - `state.db-wal` full-file sha256 at 12:05Z: `11ebff5efa4deef77c18f36332a4d3c2b6e840f86f2cf58ee4cd3178d7461254` (prefix 67,108,864 B = whole file).
+  - Historical digests `41c8be74…` (state.db) and `4a2d811b…` (container log): **Unrecovered as to byte range** — the files are being rewritten in place; a moving object's digest is not re-derivable.
+- **Live page census:** `https://alexanderwhitestone.com/stackchain/blocks/` → **121** `<li id=` entries; badges: **1 corroborated, 115 documented, 5 testimony, 0 unrecovered** (exit 0 on both greps).
+- **Repo state:** `git log --oneline -1` → `92ed02c` (reply 13, Volt's 111→95 split closure verified, grade Documented).
+- **Q53:** Closed at `f30a243` per reply 07 §3. Both residuals discharged.
+
+**Blocked:**
+- The 300s size-delta window command timed out with no output. I have before/after `stat` readings showing a 942,080 B growth in state.db over ~3 min 21 s, but no clean window measurement. This does not change the finding (the file is growing), but the exact rate within a controlled window is unmeasured.
+
+**Next:**
+- If a clean 300s window is needed for the record, re-run the delta command with a shorter sleep (e.g., 60 s) to avoid the harness timeout.
+- No other action; Q53 is closed, R5-a and R5-b are reported.
+
+**Ask:**
+- None.