Spent right? Implementing a protocol to coordinate transactions (if it is possible at all, depending on the various backends you work with) to get what you could have already done in your database?

You can keep transactions in one part of the system and the queries in another. CQRS as one example.

There are other options, like spend the same resources building a monolith that is not spaghetti. And it is certainly possible for the sum of microservices to become spaghetti.