Could take the output, hash it, and cache the result. Would allow emitting an error once (or a few times), then emitting a reminder every N occurrences.
Set an alarm and throw an exception if timeout reached before
__exit__ is called.
Possibly use an alarm to emit a logging message (or email) when time is running longer than expected.
Possibly use an average of previous run times rather than a hard-coded value.
Add option for direct logging to logstash/redis