I wish I could upvote you again. The complexity balloons when you try to adapt something that wasn't distributed, and often things can be way simpler and more robust if you start with a distributed concept.
I couldn't disagree more. My principle is to write systems extremely simply and then distribute portions of it as it becomes necessary. Almost always it never becomes necessary and the rare cases it does, it is entirely straight forward to do so unless you have an over-complicated design. I don't think I've ever seen it done well when done in the opposite direction. It's always cost more in time and effort and resulted in something worse.
Tons of vendors offer cloud first, distributed deployments. Erlang is distributed by default. Spark is distributed by default. Most databases are distributed by default.
