Are there any known issues with running a batch job that uses that State Processor APIs/Savepoint Writer with parallelism > 1? I’ve seen an issue occur multiple times when this job that is designed to run and “bootstrap” state for another job hangs in all circumstates on two keying operators unless the state is set to 1 explicitly. The job graph looks something like this:

a:(kafka-source)
 :(shuffle())
 :(map())

a:(keyA)
 :(bootstrap/transform)

a:(keyB)
 :(bootstrap/transform)

It’s similar to this issue that was reported by another user on the mailing list.