scap should catch connection timeout errors when connecting to logstash to monitor canary deploys. Currently we treat logstash failures the same as error spikes. We should at least catch the exception and report the error in a sensible way. (See T143982: scap on beta cluster does not run anymore due to logstash being down)
Scap runs /usr/local/bin/logstash_checker.py to perform the logstash canary checks and it uses its exit status to determine pass/fail. A possible fix is to arrange for logstash_checker.py terminate with a distinguished exit status if the error rate threshold has been exceeded. Other errors will result in the non-distinguished exit status of 1.
In the case of "other errors", scap will prompt the user to what to do next (exit, retry logstash check, or proceed).