[Bug] Deliberate machine disconnect emits false 'scaleDisconnected ... unexpectedly' for the machine's internal scale

[tadelv]

Summary

Disconnecting a machine through PUT /api/v1/devices/disconnect (app-initiated, clean) makes StatusPublisher emit an error-severity scaleDisconnected — "Scale disconnected unexpectedly." — for the machine's controller-owned internal scale (bengle-internal-<machineId>). The machine's own disconnect is correctly suppressed as expected; the internal scale's cascade is not.

Evidence (tablet logs, two independent occurrences)

Every deliberate API disconnect of a connected machine produced the error; a real USB detach produces both errors (correctly):

08:08:15.640 PUT /api/v1/devices/disconnect (machine usb-2e8a-a-<serial>, 200)
08:08:15.652 ScaleController - scale connection update: disconnected
08:08:15.655 WARNING StatusPublisher - emit error: kind=scaleDisconnected message=Scale disconnected unexpectedly. deviceId=bengle-internal-usb-2e8a-a-<serial>
   (no machineDisconnected error — machine expectation consumed correctly)

08:11:57.292 PUT /api/v1/devices/disconnect (machine usb-2e8a-a-<serial>, 200)
08:11:57.304 ScaleController - scale connection update: disconnected
08:11:57.306 WARNING StatusPublisher - emit error: kind=scaleDisconnected message=Scale disconnected unexpectedly. deviceId=bengle-internal-usb-2e8a-a-<serial>

By contrast, a physical USB_DETACHED at 07:58:51 emitted both machineDisconnected and scaleDisconnected — the expected behavior for a genuine unexpected drop.

Root cause

1. BengleVirtualScale.connectionState is _machine.connectionState (lib/src/models/device/impl/bengle/bengle_virtual_scale.dart) — the internal scale mirrors the machine's connection-state stream, so a machine disconnect necessarily fires the scale's disconnect in the same emission.
2. The API disconnect path (lib/src/services/webserver/devices_handler.dart, _handleDisconnect) marks only the target device as an expected disconnect: _connectionManager.markExpectingDisconnect(device.deviceId).
3. DisconnectSupervisor._handleScaleDisconnect (lib/src/controllers/connection/disconnect_supervisor.dart) consumes expectations by scale id; none was registered for bengle-internal-<machineId> — so the "unexpectedly" error fires.

So a clean, app-initiated machine disconnect always reports its own internal scale as an unexpected drop. ConnectionManager.disconnectMachine() has the same gap (marks only de1.deviceId). The sleep flow at lib/src/controllers/de1_state_manager.dart:593 does mark the scale, which is why only some flows are affected.

Suggested fix directions

• When a machine is deliberately disconnected, also mark its controller-owned scale(s) as expected — e.g. in disconnectMachine() and the devices-handler disconnect path, mark bengle-internal-<machineId> alongside the machine id.
• Or, in DisconnectSupervisor, treat a scale disconnect as expected when it mirrors a machine whose disconnect was just accepted (scale id == bengle-internal-<lastKnownMachineId> and that machine's expectation was consumed).

Repro

1. Connect a Bengle over USB (machine + internal bengle-internal-* scale become connected).
2. PUT /api/v1/devices/disconnect with the machine id.
3. Observe StatusPublisher warning error scaleDisconnected "unexpectedly" for the internal scale, even though the disconnect was deliberate.

Environment: Android tablet, Decaid with a Bengle composite USB device (machine on interface 0, serial scrubbed). Severity: low — no functional loss (the scale does disconnect with the machine), but it surfaces a false error banner/status to skins and pollutes telemetry with spurious scaleDisconnected events.