Microservices don't help polyglots make regexes faster. If you're sticking a 100ns operation behind an endpoint with JSON serialization/deserialization, network round-trip, etc., then I'm not sure what to say to that.
Actually this microservices approach has existed for decades in erlang/otp and can maintain quite excellent speed. You're not going to beat raw c speed-wise, but typically you will beat comparable dynamic languages -- and you don't have to worry about decoding/encoding json or any of that bs. I'd personally be absolutely thrilled if something like genserver and some other otp protocols were ported to other dynamic languages -- but alas real processes aren't nearly efficient enough.
A wink does not necessarily mean sarcasm? I agree that in this case it probably does, but I don't see how that use of sarcasm adds anything substantive to the conversation (aside from a not-relevant-to-this-post dig on microservices, perhaps).
On a serious note, implementing libraries as standalone applications that can be accessed via IPC is actually a viable approach. Plan 9 does this with almost all of its services. Not exactly a fair comparison though, since the IPC mechanism in Plan 9 is just the file system and much easier to learn.
