In a controlled test on an Australia PDI, I introduced a deliberate TypeError in an onLoad Client Script on the Incident form. The platform caught it and logged it at log level, not error. Live in the console, I could attribute it — the stack named the generated handler, and the rendered page had the mapping back to the script name and sys_id. After closing the tab, all I could find was the timing record in syslog_client_trans_det — type, duration, order, a shortened name. No exception message, no stack, no failure flag. Searching syslog for the exception text and probe name returned no matches. For people running real instances: when a form breaks for one user and you can't reproduce it, what native tools or retained evidence would you check next?
