Host Crash Loop + Error Reporting

It appears that in some cases when the runtime is in a crash loop (e.g. due to an incorrectly configured ServiceBus trigger connection string) the portal will sometimes stop showing the actual underlying error and will instead show a generic 500 error with no details.

Repro steps:

- Create a ServiceBus queue triggered function for a queue that doesn't exist
- set the connection to use a SB connection that doesn't have Manage rights

The host will enter an error loop in this case, because it is trying to create the queue but doesn't have rights. In the portal, initially you'll see the correct error message, but if you dismiss it and wait for a couple minutes, you'll start getting the bad error message (just generic 500).

Reproduce the Service Bus trigger scenario with a nonexistent queue and a connection lacking Manage rights, then observe the host and portal behavior during the crash loop. Trace the host/runtime error-reporting path involved after the initial error is dismissed; done means the underlying error remains available instead of becoming a generic 500 response.

Written by the indexing model from the issue text.