I've found the exact opposite. We have lots of gcp cloudsql instances. Each team handles their own dbs for however they decide to split their services.

Our operational costs are more distributed, they are more but they indicate the cost of each product and they separate our data very well.

Teams can do their own migrations, and we can prevent some rogue service from violating the data.

We are currently in progress for a very large migration from what you have. But I'm sure we just did it wrong.