Matt Butcher

Incidentally, using the `--release` flag on the Grain compiler seems to have fixed this issue. I am guessing that this may have to do with de-duplicating imports

There is some work on this in the SMI trait, but I agree we should do a Deployment-based version.

For now, it would be fine to do it ourselves. Clux might be interested in adding something to his library if we wanted. Or if you would like to write...

Relevant `nodeSelector` docs: https://kubernetes.io/docs/concepts/configuration/assign-pod-node/

Can you point me to docs? I'm not totally sure what I am looking for.

Do we have milestones for the spec that will require Rudr rewrites or new Rudr features?

Do we want this before the July release? It is not in parity with the managed service.

I've started on this one, but it requires upgrading our core Kubernetes library, as well as some additional setup. So it will take me a while. Depends on #22

Split is the only high-priority one, and I'd like to implement a Canary trait on top of that.

I think that what we want is for the number of completions to be equal to the number of parallel instances. But the problem is that the `parallelism` field is...