Hi there channel! I have a PR in which I’m making a couple of suggestions which I’ll describe in a couple of threads. The PR seems to stabilise our system, where ideally we would have an arbitrary number (~10) of concurrent flows with an arbitrary of batch workers (max-workers~100 per task, max-num-splits~1000). At this scale (which I wouldn’t say is enormous) we find that we hit transient errors on a regular basis (~once per week) and we’re only a small team so we’d like to limit this as much as possible. First part of the PR, as suggested in the comment of another PR makes the throttle parameters configurable. As savin has described in the PR, which cautiously calls into question the necessity of the feature:

given that AWS Batch is a managed service, I would expect almost all the users to rely on the default values that we would configure

which I agree with, but I’d add the counter that if it makes no difference to most users and provides a transparent mechanism to stabilise production systems then it seems useful nonetheless. (it also might be useful to implement this as a feature on it’s own) Potential argument against this feature might be (I can’t see where this would come into play though) that if the user decides to set the polling threshold 

BATCH_DESCRIBE_JOBS_THROTTLE_DELTA

 too high that could lead to some side effect that I’ve not considered

@dry-state-56742 re: making the values configurable - as I mentioned in the gh comment - let's make it configurable, with the expectation that most people will never override it (the defaults should work for most of the users).

ok great, in which case I’ll open a PR for this shortly