Good diagnostic logging layer on top of the #1960 policy extraction. The rate-limited waterfall decision log with state-change detection is well thought out, and including XVTR state in the troubleshooting snapshot will save real debugging time. A few observations:

static QHash in logXvtrWaterfallDecision — thread safety

MainWindow.cpp around line 241: logStateByStream is a static QHash inside a free function. Since waterfall tile callbacks arrive on the VITA-49 stream thread (not the GUI thread), this static is mutated from at least two contexts without synchronization. In practice today this is likely fine because the lambda capturing the call site runs on the main thread via QueuedConnection, but if the connection type ever changes or this function gets called from a background path, this becomes a data race. Worth a brief comment noting the single-thread assumption, or switching to a member variable on MainWindow to make the lifetime/thread ownership explicit.

xvtrPolicyBandsFrom allocates on every waterfall tile

The conversion xvtrPolicyBandsFrom(m_radioModel.xvtrList()) is called inside the waterfall tile lambda (line ~1721) and again in the band-switch path (line ~8271). In the waterfall path this runs per tile per panadapter match. For the typical case (0-3 XVTRs) this is negligible, but worth noting — if someone configures many XVTRs and has multiple waterfall streams, the QVector allocation per tile adds up. Not a blocker, just something to be aware of for future optimization if profiling shows it.

formatXvtrBullet — QString::arg with >9 positional args

SliceTroubleshootingDialog.cpp line ~108: The format string uses %10 which with Qt's QString::arg chaining will work correctly here because the args are applied sequentially. Just noting for future maintainers that this is at the edge of arg()'s positional numbering — adding an 11th field would need the multi-arg overload or split formatting.

hasIsValid field addition — good

The hasIsValid field on XvtrInfo (RadioModel.h:574) cleanly distinguishes "radio says is_valid=0" from "radio never sent is_valid". This is exactly the kind of diagnostic metadata that was missing. The sticky-true semantics (x.hasIsValid = true once seen) are correct for a session-scoped flag.

Scope check

All files are within the stated PR scope (XVTR diagnostic logging + snapshot). CMakeLists.txt changes are limited to adding the new source and test target. No out-of-scope modifications detected.

Tests

The test target follows the existing passive_spots_policy_test pattern (standalone, no CTest registration, Qt6::Core only). Consistent with the project approach.

Overall this is a clean logging-only addition with good rate limiting and structured diagnostic output. Thanks for the thorough PR description and validation steps, @jensenpat.