Without notifications working in on-prem I don't think you need to spend effort on alerts if deployments fails. Generally speaking you're either launching it from the UI and can easily see the status, or launching with automation which can tail the request and alert if required. It's obviously not ideal but with how early/unfinished the product is now you may spend a lot of time iterating on something that gets thrown away with the next minor release.
As for failure states in debugging the basic build, I currently use EBS recovery runnables. Using certain properties in the payload I can either skip all EBS subscriptions, or if my custom property "debugModeEnabled" is true and the primary EBS target fails, the recovery runnable checks that property. If true don't error, if false throw the original exception; in both cases skip all remaining EBS.
