Field notes — curtain_motor "doesn't move" + mmWave class flip (2026-10-10)
curtain_motor: _TZE200_127x7wnl, _TZE204_5slehgeo, _TZE200_icka1clh
Sources: Homey diagnostics (read-only Gmail, [email removed]), logs ab5aaf04 (01/09, 9.0.775) and the 15-log thread 11/09–13/09 (48baba36 … ed0de063, app 9.0.874–9.0.908); Z2M zigbee-herdsman-converters src/devices/tuya.ts (Koenkk & contributors) for icka1clh.
What the logs show (rewritten, not quoted):
- 9.0.775:
[TUYA] Failed to send DP1/DP8: Tuya cluster not availablewhile the TX wrapper still logged success → false success, motor never moved. Fixed by P2380 (fa31ad8575, 02/09) + P2403 soft skip (4e699672c4, 03/09). - 9.0.874–9.0.908 (5slehgeo / 127x7wnl, "Still not responding"): EF00 is detected, Moes ZTS path sends DP1+DP2 with inverted position, Homey reports TX OK, but the device never sends anything back; one TX ends with Homey "device unreachable". Fixed afterwards by P2467 (f575a24fcd, 11/09: real mcuSyncTime 0x24 + EF00 init before first motion), plus single-retry for Moes ZTS (P2480).
- A sibling cover that does report (DP2 119 → 100, DP1 1) shows the RX path works, but each frame is processed ~6× (duplicate listeners, suppressed by dedupe).
- icka1clh is Z2M
TS0601_cover_4(Moes AM43-0.45/40-ES-EB): DP1 control, DP2 set position, DP3 reported position, DP5 direction, 101 mode, 105 speed; no ZCL 258. Already EF00-only in our ZCL fallback guard (P2461). - 127x7wnl / 5slehgeo are NOT in Z2M; the "Moes ZTS-EUR-C" label comes from our GitHub #533.
Verdict: every failure pattern in the reports predates fixes already on master, stable-v5 and bastien-home; no report since 13/09. No new runtime change made.
Ask for a new diag if a user reports it again (app ≥ current): the log must show [CURTAIN] P2467 Moes MCU time-sync armed, the P2412/P2464/P2467 Moes ZTS TX line, and 30 s after one Open press whether any DP… received line comes in; plus the device interview (endpoint clusters) and whether the wall button moves the motor. No RX + unreachable = mesh/range; RX but no motion = DP mapping/invert issue.
Open: RX listener fan-out (~6× per frame) on curtain_motor — harmless today, audit later.
mmWave _TZE200_3towulqd class change
Only in presence_sensor_radar (class sensor) on all three branches; device.js refuses any setClass other than sensor (P2548, f509fb2e83, 17/09). Reports through 21/09 are on builds around that fix. Nothing more to change.