Auto: ablegacy-9 — per-device timeout / retry overrides

Closes #252

tests/ZB.MOM.WW.OtOpcUa.Driver.AbLegacy.IntegrationTests/Docker/README.md

@@ -139,6 +139,38 @@ bit writes to real hardware or RSEmulate 500 until upstream resolves.
 See [`docs/drivers/AbLegacy-Test-Fixture.md`](../../../docs/drivers/AbLegacy-Test-Fixture.md)
 for the full coverage map.
 
+## Per-device timeout fixture (PR 9 / #252) — TODO
+
+`AbLegacyPerDeviceTimeoutTests.Per_device_Timeout_below_simulated_delay_surfaces_BadCommunicationError`
+needs a slow-link sidecar before it can run for real. The simulator answers
+in <100 ms locally, so a 500 ms per-device timeout never trips against
+the unmodified container.
+
+Two options, neither wired up at PR 9 time:
+
+1. **`tc qdisc` netem inside the container** — add to `docker-compose.yml`:
+
+   ```yaml
+   # services:
+   #   ablegacy-slc500-slow:
+   #     extends: ablegacy-slc500
+   #     cap_add: [NET_ADMIN]
+   #     command: >
+   #       sh -c "tc qdisc add dev eth0 root netem delay 800ms &&
+   #              ab_server --plc=SLC500 --port=44818 --path=1,0 --tag=N7[200]:INT16"
+   ```
+
+   `--cap-add=NET_ADMIN` is required because `tc qdisc` mutates the
+   container's egress queue. Combine with `AB_LEGACY_COMPOSE_PROFILE=slc500-slow`
+   to point the suite at the slow profile.
+
+2. **`iptables --delay` shim** — sidecar container that NATs port 44818 and
+   adds a fixed delay on the SYN/ACK + payload path. More portable than
+   netem (no `NET_ADMIN` on the simulator itself) but adds a hop.
+
+When either lands, drop the `Skip = …` on the integration test and assert
+the precedence rule end-to-end.
+
 ## References
 
 - [libplctag on GitHub](https://github.com/libplctag/libplctag) — `ab_server`