There are two problems with this approach.

One: the amount of manual analysis that needs to go into deciding whether certain conflicts are tolerable or not is unmanageable for even mildly complicated use-cases, so you can't know without either formalising a TLA+ model of your _whole_ distributed state or checking potentially millions of combinations.

Two: The 'no true Scotsman' fallacy. If not having strong coupling between your microservices is the definition of what a microservice is and given enough complexity all microservice projects develop problems around this, then it follows that there are no microservice projects in the world.

This is what my next article will be about.