Is there any retry mechanism if the batch fails?

There is a built-in assumption here that any failure of a single batch can be ignored as it gets retried frequently. I suppose consecutive failures should surface some form of error somewhere. I wonder if the database health page could show a summary of queries that have failed multiple consecutive executions?

There is a built-in assumption here that any failure of a single batch can be ignored as it gets retried frequently. I suppose consecutive failures should surface some form of error somewhere

I might have missed it, but it would be cool to leave a comment stating this.

Done

wat

Nice strictness check! Feels like this could be a testutil func, even if obvious.

If we need to do this instead of just returning, I'd prefer returning an error here, wouldn't want this to happen for a production instance after some changes to the code.

Feels like this mostly exists for tests (otherwise could be launched fromNew), so more reason not to add a possibility of panic.

If we return an error, then we have two options:

1. Exit with the error, which brings us back to potentially impacting prod, but just in a nicer-looking way.
2. Just log the error and do nothing, which means the error will likely pass unnoticed until folks wonder why their workspace usage isn't being tracked.

I don't think 2) is a valid option. We're not guaranteed to catch it in tests especially because of the possibility of ignoring errors inslogtest.

There's an alternative possibility 1a) where we check the error and recreate the tracker if non-nil. WDYT?

I forgot about option c) - haveNew start the loop and unexport the function.
Now it's un-possible!

Lol, love the personality in this comment 😄