Fix eight review findings on the render-bounds diagnostics

Corrects a false comment example, fixes two tests that couldn't detect their
own regressions, adds two more read-back checkpoints around Main_OnCommand so
a drift report self-locates, guards a spurious zero-vs-zero coincidence match,
and softens two sentences that overclaimed cause or defect.

src/shell/capture/capture_realtime_shell.cpp

@@ -369,6 +369,11 @@ RealtimeRecordBackend::begin(const CaptureRequest& request,
 
     // recordWindowEnd extends past the range end for a tail mode so the
     // transport captures the decay; cursor + time selection are restored by restore().
+    // `[verify — DAW]` whether rs/re come back changed on this isSet=true call: the SDK
+    // header names both `double*` but documents no read-back semantics for either
+    // direction, and nothing here reads rs/re again after the call to notice. Lower
+    // stakes than the offline RENDER_* store: completion is driven by the play cursor
+    // reaching the range end (tick(), below), not by re-reading this pair.
     double rs = request.startSeconds, re = st->recordWindowEnd_;
     GetSet_LoopTimeRange(true, false, &rs, &re, false);
     SetEditCurPos(request.startSeconds, false, false);