We have some custom email commands for which this is particularly important, eg, we have an analogue to landing a diff available to us as an email command. For this email command, and others, the best you can do when an email command fails currently is just "do nothing at all and hope the user eventually figures out that this command did nothing". This is particularly bad if you were counting on the email command to do something for you like land your code.

Ideally email commands would provide error handling mechanisms which respond to the sender in the case of failure.