Clean, minimal fix. The model-owned m_radioTransmitting flag is correctly stored alongside the existing radioTransmittingChanged emit, reset on disconnect, and used to gate the P/CW compression gauge through the meter handler. The explicit updateCompression(0.0f) on the TX→RX transition is a sensible belt-and-suspenders against quiescent RX meter publishes that would otherwise leave the gauge pegged until the next mic-meter event arrived.

A few observations:

1. Order of operations is correct: in onStatusReceived you set m_radioTransmitting = radioTx before emitting the signal, so any handler reading isRadioTransmitting() synchronously sees the new value. Same pattern in onDisconnected.

2. Access pattern in the new TX→RX clear is consistent with the surrounding lambda — m_appletPanel->sMeterWidget() is dereferenced two lines above, so the new phoneCwApplet() call has the same lifetime guarantee. No null risk introduced.

3. One thought, not blocking: since the micMetersChanged lambda also gates on isRadioTransmitting(), the explicit clear on the radio-transmitting signal is only load-bearing when the radio stops publishing mic meters during RX. Worth a one-line comment on that explicit clear so a future reader doesn't think it's redundant with the gating below.

4. Out of scope but flagged in the PR body: noting meter_model_test failing on main independently — agree, that's not yours to fix here.

Thanks for the focused fix and the clear root-cause writeup, @jensenpat.