Language worker "User" logs are ignored after execution is complete

Back when @yojagad added a LogCategory for language workers to pass either System or User in this PR to the node worker, @pragnagopa said

With in the execution context, we should not be logging system logs

However, some of the logs in question are like this:

Error: ‘done’ has already been called. Please check your script for extraneous calls to ‘done’.

Which actually happens after execution completes. As of today, those logs are silently ignored by the host here (nothing in app insights, nothing in the console, etc.). Which leads me to a few questions:

• What exactly is the difference between System and User logs? (We might want to document in the protobuf definition)
• How should language workers categorize logs that are related to an execution, but after an execution?
• Should the host really be ignoring these logs? These logs could be coming directly from the user, and perhaps they don’t realize they log after execution completes because they never see the logs.

cc @alrod

[ejizba]

Logs between invocation request and invocation complete should be considered user logs

If the language worker is crafting the message itself, can’t we consider it system logs regardless of when it happens? There’s no risk of PII as long as it’s a simple hard-coded string, which is why I assume it’s okay for the host to write system logs during this time. Having those worker logs in our telemetry would be very helpful for supportability of language workers.

[fabiocav]

@AnatoliB it would be good to sync on this to make sure we’re not duplicating anything you’re already planning for this area.