Demo (SimBackend) doesn't model the radio-side autopan recenter — panadapter band-follow is unobservable in the simulator

Summary

On real Flex hardware, a band-change tune (slice tune <id> <mhz> issued without autopan=0) makes the radio recenter the panadapter on the new frequency and report the new center back to the client. The demo SimBackend does not emulate this: it moves the slice but never moves the pan center. As a result, the panadapter's cross-band band-follow behavior cannot be observed or tested in demo mode — only on hardware.

Observed vs. expected

Evidence (measured via the automation bridge)

• Demo: set_freq 14.1 → 7.1 → get_freq = 7100000, but get_state pan centerMhz stayed 14.1.
• Real 6500: identical command → pan centerMhz followed 14.261 → 7.1 (a panadapter screenshot showed the 40m band-plan and the slice cursor at 7.1). Same on the SmartCAT-Flex and TS-2000 CAT planes (→ 21.1, → 28.4).

Root cause

• SimBackend::setSliceFrequency() (src/core/backends/sim/SimBackend.cpp) updates m_sliceFreqMhz and emits sliceChanged, but never touches the pan center. A setPanCenter() / panCenterBandwidthChanged path exists in the same file but isn't invoked on a slice tune.
• The typed backend seam (RadioConnection → SimBackend::setSliceFrequency) also strips the autopan flag, so the sim can't distinguish a recentering tune from an autopan=0 one even if it wanted to.

Impact

CI and demo-mode testing can only assert the CAT-visible frequency, not the pan-follow — which is the actual behavior of the CAT band-follow feature (RFC #4333). The CAT test suites' §2c already acknowledge this ("pan-follow is CAT-invisible; confirmed visually"). Anyone using the demo to evaluate or develop band-follow will see it appear "broken" when the code is in fact correct — pan-follow currently can only be validated on hardware.

Suggested fix

Have SimBackend model the autopan recenter: on a band-change tune (out-of-span, i.e. no autopan=0), recenter the pan on the new frequency and emit panCenterBandwidthChanged, mirroring how the radio behaves. This needs the seam to carry the autopan flag (or SimBackend to infer out-of-span the way RadioModel::tuneSliceForCat does). Then demo mode + CI could exercise band-follow end-to-end.

Steps to reproduce

1. Launch AetherSDR, connect to Demo.
2. Note get_state pan center and the slice frequency.
3. Cross-band retune via any CAT port (e.g. rigctld set_freq to a different band).
4. Observe: the slice frequency changes; the pan center does not.
5. Repeat on real Flex hardware → the pan center follows.

Related sim-fidelity gap (possibly separate issue)

The demo also doesn't model a radio-side CW keyer: SmartCatProtocol::cmdKY returns ?; when !hasRadioSideCwKeyer(), which is the case in demo mode, so the TS-2000 suite's 12.13 KY; check fails against the sim (it asserts KY0/KY1). Same class of gap — the demo doesn't emulate a radio-side capability — noted here for awareness; may warrant its own issue.

🤖 Reported with assistance from Claude Code

Start with src/core/backends/sim/SimBackend.cpp and trace the RadioConnection seam into SimBackend::setSliceFrequency(). Compare that path with RadioModel::tuneSliceForCat and the existing setPanCenter()/panCenterBandwidthChanged handling. Done means demo cross-band tunes reproduce radio pan-follow while autopan=0 and in-span tunes retain their current behavior, with the relevant CI or CAT coverage updated.