docs: FEAT-14 — second T3U to replace onboard wlan0 (dual AC radios)

FEATURES.md

@@ -33,6 +33,7 @@ dedicated PLAN.md phase (or fold it into the relevant existing work) and mark it
 | FEAT-11 | `SHIPPED` | Hotspot SSID/password settings + 5GHz WiFi role swap | Phase 27 |
 | FEAT-12 | `SHIPPED` | Log OBD speed before GPS fix (rolling-start capture) | Phase 27 |
 | FEAT-13 | `IDEA` | Flush uploaded tracking rows from local DB | unassigned |
+| FEAT-14 | `IDEA` | Second T3U to replace onboard wlan0 (dual AC radios) | unassigned |
 
 ---
 
@@ -256,3 +257,21 @@ dedicated PLAN.md phase (or fold it into the relevant existing work) and mark it
     urgent — imported dashcam videos are the real disk consumer.
 
 
+### FEAT-14 — Second T3U to replace onboard wlan0 (dual AC radios)
+- **Status**: `IDEA`
+- **Area**: hardware + `deploy/yaris_uplink.py`, `deploy/install_netwatch.sh`, `modules/app_network.py`
+- **Phase**: unassigned
+- **What**: Owner bought a second TP-Link Archer T3U (RTL8812BU) to replace the onboard
+  BCM43438 currently serving as the internet uplink (wlan0). Both radios would then be fast
+  AC1300 T3U dongles — one AP (wlan1, 5GHz) and one uplink (wlan0) — removing the current
+  tradeoff where away-from-home uploads run over the weak onboard radio.
+- **Considerations**:
+  - Both dongles share the Pi 3's USB 2.0 bus (~480 Mbps total), so two AC dongles can't both
+    run at full speed simultaneously — but each still far exceeds the onboard radio.
+  - Power: two AC dongles draw more than onboard + one; verify the 5V GPIO feed holds
+    (`vcgencmd get_throttled` should stay `0x0` for no undervoltage).
+  - Interface naming: both radios are USB now, so `wlan0`/`wlan1` may not map predictably —
+    pin `connection.interface-name` and/or add MAC-based udev rules so AP vs uplink never swap.
+  - The rtl88x2bu driver is already built; a second identical dongle should be plug-and-play.
+
+