I’ve always thought the monolith vs. micro service debate to miss the point. Taking an RPC call has costs. But sometimes you need to. Maybe the computation doe

I’ve always thought the monolith vs. micro service debate to miss the point. Taking an RPC call has costs. But sometimes you need to. Maybe the computation doesn’t fit on any of your SKUs, maybe it is best done by a combination of different SKUs, maybe you need to be able to retry if an instance goes down. There are countless reasons that justify taking the overhead of an RPC. So, do you have any of those reasons? If yes, take the RPC overhead in both compute and version management and stuff. If no, don’t. Why is this so contentious?