On Linux, a significant amount of the host functionality is implemented by a Python wrapper script that currently has no way to report failures other than logging to a local file. It would be useful if that script could report certain classes of failure to the directory, so that it would be more visible to the user. Two specific examples:

* A user's session script exits prematurely.
* The host process crashes beyond the restart limit.

In either case, we can report a specific failure reason to the directory, which could then have a suitable error message displayed on the client.