This probably is the reason why it works in some settings. Anecdotally, as someone with more domain/functional knowledge and operations knowledge than knowledge of frameworks, I found microservices architecture with good functional test coverage a better way to deal with a team of programmers like me (basically, a typical team in an offshore IT consultancy building enterprise applications using SpringBoot and Node.JS). Ship the service as early as possible with available talent and then get someone really good with that programming language or framework to deal with performance bottlenecks within the microservice. I see it basically as a way to limit the blast radius of the applications. Of course, as you said, get the boundaries wrong and you have a bigger problem.