diff --git a/BUGS.md b/BUGS.md
index 61012b7..d849859 100644
--- a/BUGS.md
+++ b/BUGS.md
@@ -67,7 +67,12 @@ the authoritative, issue-by-issue record lives **here**.
 ---
 
 ### BUG-03 — DTC/MIL codes are never persisted; lost on OBD disconnect or reboot
-- **Status**: `OPEN`
+
+- **Status**: `FIXED` (commit `e2f2314`; pending deploy/verify → CLOSED)
+- **Fix**: Persist MIL + codes to SQLite on each poll (`save_dtc_state`), `/api/dtc` falls back to
+  the persisted snapshot when OBD is offline, `get_overview_stats` counts the real `dtc` table,
+  and a schema migration rebuilds the unused `NAME/STATUS/DTC_ID` table. Cleared codes are tracked
+  via `IS_SET` + `FIRST_SEEN_AT`/`LAST_SEEN_AT`. Tests: `tests/test_dtc_persistence.py`.
 - **Severity**: Medium (lost diagnostic data; also a stats bug)
 - **Area**: `modules/app_obd.py`, `modules/app_sql.py`, `get_overview_stats`
 - **Phase**: Phase 26
@@ -101,7 +106,6 @@ the authoritative, issue-by-issue record lives **here**.
   two systems agree. Note this may re-split/recompute existing server trips — get owner sign-off
   before recomputing, and re-run `scripts/recompute_trips.py` after the fix.
 - **Acceptance**: A real drive with a parked gap produces the same number of trips on the server
-  as on the Pi.
 
 ---
 
diff --git a/PLAN.md b/PLAN.md
index 39c6225..f72244c 100644
--- a/PLAN.md
+++ b/PLAN.md
@@ -1876,7 +1876,8 @@ lands. The current known issues from the road test (also recorded in `BUGS.md`):
    unicast fallback can be extended to the AP subnet.
 2. **BUG-02 (micro-USB power port damaged).** Hardware-only (worn power connector). Action is
    documentation + owner hard-wiring 5V/GND to the GPIO header pins; no code.
-3. **BUG-03 (DTCs not persisted).** The SQLite `dtc` table is created but never written or read;
+3. **BUG-03 (DTCs not persisted).** ✅ FIXED (`e2f2314`) — The SQLite `dtc` table is created but never
+   written or read;
    DTC codes/MIL live only in memory in `app_obd.py` and vanish on OBD disconnect/reboot.
    `get_overview_stats` also queries a non-existent `dtc_codes` table. Fix: persist on each poll,
    read back when disconnected, fix the stats query, add a "cleared/since" concept.
@@ -1884,7 +1885,8 @@ lands. The current known issues from the road test (also recorded in `BUGS.md`):
    speed + sustained GPS displacement, inactivity gap); the server splits only by *time gap* and
    filters by duration. Parked idle tracking rows bridge two drives into one server-side trip.
    Fix: mirror the Pi's motion-boundary logic server-side.
-5. **BUG-05 (server replay speed_source).** Server `get_trip_route` hardcodes `speed_source: "gps"`
+
+5. **BUG-05 (server replay speed_source).** ✅ FIXED (`e2f2314`) — Server `get_trip_route` hardcodes `speed_source: "gps"`
    even when OBD speed is present (it already returns raw `speed`). Fix: mirror the Pi's
    `obd if speed>0 else gps` logic.
 6. **BUG-06 (server timezone is UTC-shifted).** `tracking_points.recorded_at` is `TIMESTAMPTZ`
@@ -1900,6 +1902,13 @@ lands. The current known issues from the road test (also recorded in `BUGS.md`):
    *server* delete endpoint, and there is no `DELETE /api/trip/<id>` route on the Pi. Fix: add all
    four pieces and make the delete endpoint configurable in `history.js`.
 
+**Progress (2026-08-11):** **BUG-03** ✅ FIXED (commit `e2f2314` — DTC/MIL persisted to SQLite on
+each poll, `/api/dtc` falls back to persisted state when OBD is offline, `get_overview_stats`
+fixed to use the real `dtc` table, cleared/since tracked via `IS_SET`/`FIRST_SEEN_AT`/
+`LAST_SEEN_AT`, plus a schema migration and `tests/test_dtc_persistence.py`). **BUG-05** ✅ FIXED
+(commit `e2f2314` — server route `speed_source` mirrors the Pi's `obd if speed>0 else gps`).
+Both pending deploy/verify. The remaining items (BUG-01, 02, 04, 06, 07, 08) stay open.
+
 **Acceptance for Phase 26:** every `BUGS.md` entry in scope is fixed, regression-tested, deployed,
 marked resolved, and the corresponding fix recorded here.