Imagine the goal is to fix the problems (e.g. make it possible to run less of the services or something like that): How do you do that without first running all the services, making the proper changes, and then testing those changes? You need to be able to run all the services in that interim period.

So, wouldn't it be nice if there were a solution for this in-general? And, maybe, it would lead to better conditions later on. But in the meantime there is really no way around the existing design/decisions/etc. You simply have to deal with that reality and engineer around it.