DataFlow max worker parallelism

[Joey Tran]

Out of curiosity, is there a technical max number of workers that the
DataflowRunner can run? Or is there no theoretical limit, so long as you
horizontally scale the runner node?