Bodyless red-log captures no longer become undiagnosable incidents

The red-log ticketing pipeline could open an incident — and a GitHub issue — whose entire content was a console header such as OrleansRoutingService[0]: no message, no exception, no stack trace, nothing an engineer could act on. Every later bodyless capture from the same component then folded into that one bucket instead of being noticed. The watcher that reads the logs had been fixed to hold such lines back, but the watcher is a separately shipped component and the running one predated the fix.

The portal now checks every incoming report itself. A report that carries no diagnostic is refused and recorded on the per-namespace "capture gap" finding — the same one an up-to-date watcher files — with the original component and fingerprint kept as evidence, so the incident list shows where bodyless captures are coming from rather than a ticket nobody can diagnose.