Cache and requests: Handling of dead status endpoints (Internal Server Error, old requests)

The requests to the TOAR database indicate, that an internal server error is reported as soon as the analysis service crashes.

Accessing the status endpoint afterwards will simply return the status endpoint without any further progress.

Possible solution: as soon as an internal server error is reported, aboard the current request and start a new request. This requires a careful handling of the different errors, especially connection errors need to be separated.

An addition adoption might be to save the creation date of a request into the cache and discard saved endpoints after an adjustable duration.