A problem I see with many Microservice vs Monolith discussions is that there seems to be an assumption that it is a binary relationship instead of an analog one

A problem I see with many Microservice vs Monolith discussions is that there seems to be an assumption that it is a binary relationship instead of an analog one. That is, there exists all sorts of architectures in between the microservice model and monolith model that have different trade-offs. You have programs like postfix which has a collection of small applications working together via sockets, queue based separation like sidekiq, RPC systems like protobuffers or thrift, etc. You don't have to use an HTTP/REST API to get many of the benefits of microservices while controlling the complexity.