Keep the monolith, but split the workloads
I’m a big fan of monolithic architectures. Writing code is hard enough without each function call requiring a network request, and that’s before considering the investment in observability, RPC frameworks, and dev environments you need to be productive in a microservice environment.