docs(omny): note readoutPriority: on_request was kept deliberately

Mirko confirmed keeping on_request (not reverting to the original async)
even though it turned out not to be what fixed the omny_fermat_scan crash
-- these cameras are only ever used for live GUI viewing, never read back
in scan code, so async's continuous per-scan logging was likely unused
anyway. Recorded so this doesn't read as leftover cruft later.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01QLrD7sVYGLAzsQjLVJpCgt

csaxs_bec/bec_ipython_client/plugins/omny/AI_docs/OPEN_ISSUES.md

@@ -187,6 +187,16 @@ crash, and `scan_actions.py`'s `_send()` should scope the `response` flag per-de
 per-batched-instruction -- any scan with this same signal-vs-root-device mismatch pattern would hit
 the same failure mode.
 
+**Note on the earlier `readoutPriority: on_request` change (commit `c37bfc7`)**: once this fix
+landed, that change was no longer load-bearing for this crash -- the rotation's own readback
+progressbar now closes out right after the move, so it's no longer still listening when
+`complete_all_devices()` broadcasts later, regardless of cam200-203/cam_xeye's readout priority.
+Asked Mirko whether to revert it back to `async` (original behavior, cameras' frames get
+continuously logged into every scan) now that it's not strictly needed; he chose to **keep
+`on_request`** -- these cameras are only ever used for live GUI viewing (`start_live_mode()`), never
+read back in any scan code, so continuous per-scan logging was likely never useful. Not reverted;
+this is a deliberate choice, not leftover cruft.
+
 ## Environment note (not code, but will bite again if forgotten)
 
 Both `csaxs_bec` and `bec_widgets` were pip-installed editable pointing at pre-repo-