We are currently making every Docket proactively create the docket stream and worker group (as well as maybe the strike stream too?) on initialization, but that's only necessary for the rare first time that you're using a given docket. After that, it's just a waste. The goal here was to avoid errors when you .add() and so on, but the runtime penalty may be too high.
Let's evaluate an approach where we can recover and retry when the streams we need aren't there.