…ciple II.

TransmitModel::setSbMonitor sent "transmit set mon=" but never updated its
local m_sbMonitor, so transmitModel().sbMonitor() read stale until the radio
echoed sb_monitor (measured non-convergent over a 410 ms window on a
FLEX-8400M). Same defect class as #3712 (setAmCarrierLevel).

Mirror the sibling pattern: write m_sbMonitor optimistically on change and
emit micStateChanged() — the signal the MON button's model->widget sync
(PhoneCwApplet::syncPhoneFromModel, connected to micStateChanged) actually
binds to, matching setMonGainSb. NOT phoneStateChanged() (that drives
syncCwFromModel, which never touches the MON button). No feedback loop: the
sync wraps its body in m_updatingFromModel and the toggled handler early-
returns on that flag, so the optimistic setChecked cannot re-emit the command.

Principle II: the command path stays client->radio; the radio's sb_monitor
status echo (parsed at TransmitModel.cpp:104-106) supersedes the optimistic
value, with no feedback loop. Command/status split per Principle I (FlexLib:
command "mon", status key "sb_monitor", Radio.cs:10615).

Blast radius: risk_score=0.411, 23 high-risk affected (top:
MainWindow::MainWindow, MainWindow::wireRadioModel, RadioModel::
troubleshootingSnapshot). All structural noise from class-level granularity;
the change only adds an existing signal emit already wired and feedback-
guarded, no signature/symbol/dependency change. Test-target link surface
unchanged (no new external symbol or include). Note: cmake unavailable in this
environment, so a full build was not run; the edit uses only pre-declared
members/signals isomorphic to verified sibling setters in the same TU.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

…inciple II.

setCwSidetone() sent the `cw sidetone` command but never updated its local
member m_cwSidetone, so cwSidetone() returned a stale value until the radio's
status echo arrived. Apply the canonical guarded-optimistic pattern (identical
to #3712 setAmCarrierLevel and the adjacent setCwBreakIn/setCwDelay/setCwIambic
setters): update m_cwSidetone and emit phoneStateChanged() when the value
changes, before sending the command.

No feedback loop: PhoneCwApplet's button->model edge is guarded with
!m_updatingFromModel and its phoneStateChanged sync wraps widget updates in
m_updatingFromModel=true/false plus QSignalBlocker, so the optimistic emit
cannot re-send the command.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>

TransmitModel::setDax sent the radio command but never wrote its local
m_daxOn member, so a read in the same tick reflected the change only
after the radio's dax= status echo (~80ms live). This mirrors the
sibling mic setters (setSpeechProcessorEnable, setMicLevel, setMonGainSb)
which update optimistically and emit micStateChanged(); setDax was the
odd one out. Latent today because the radio echoes dax= promptly, but a
slow or dropped echo would surface the same stale read as #3712.

Emits micStateChanged() (not phoneStateChanged() from the issue body):
the echo path sets m_daxOn under the micChanged flag and emits
micStateChanged() (TransmitModel.cpp:232), and PhoneCwApplet syncs the
DAX button in syncPhoneFromModel() (PhoneCwApplet.cpp:740), wired to
micStateChanged (:667-668). The re-emit is safe: syncPhoneFromModel()
sets m_updatingFromModel and wraps the button in QSignalBlocker, and the
toggled handler is gated by !m_updatingFromModel (:313-315).

commandReady stays unconditional, so callers that depend on the command
always reaching the radio (Ax25HfPacketDecodeDialog save/restore,
MainWindow shortcut/controller bindings) are unaffected; the optimistic
read actually improves the daxOn()-then-set call sites.

Blast radius: risk_score=0.128, 6 high-risk affected (top: MainWindow::MainWindow, Ax25HfPacketDecodeDialog::Ax25HfPacketDecodeDialog, MainWindow::registerShortcutActions). All real call sites verified non-regressing; high-risk entries are wiring/construction bridges that only instantiate TransmitModel.

Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>