Completely agree. This is also almost the same as the distributed monolith diagnosis.

There’s also this new mind virus I’m seeing in the industry: the “orchestrator” service. This almost always is a bad idea.

Orchestrators only work when you working in a completely different substrate like big data (spark) or online inference.

But as a normal service should never be an orchestrator because by definition, useful changes spread across its children nodes. But I have no idea who it became popular in enterprise.