We use flows triggered by process builder, triggered by platform events and we test them using external tests written in Python.
We know the processing is asynchronous and we correctly implemented the retry logic in our assertions (queries on SObjects, typically), but it seems sometimes those events are not correctly processed and the failures are intermittent. We manage to reproduce the failure “sometimes” against a newly created scratch org. This leads leads two questions:
- can platform event publishing fail on a newly created scratch org?
- is there some sort of lambda that needs to be warmed up, and have high spikes in the lambda warm up times?