Guys. Just hear me out. What if we will build a microservice mesh for API calls runtime validation, instead of designing robust API, huh? https://drops.dagstuhl.de/storage/00lipics/lipics-vol194-ecoop2021/LIPIcs.ECOOP.2021.20/LIPIcs.ECOOP.2021.20.pdf I am kidding - overall the paper is quite interesting, and the approach as a crunch, but useful and robust crunch might be useful for improving legacy systems, I think. The biggest questions are: 1. How to expresivelly define these types? It feels like an openAPI addition with the definitions of call sequences. Which is already bloat as hell. 2. How the monitor agents will behave if we will specify a dozen on session types, we want to validate? I think the overhead will be worse than the one they measured in the paper on kindergarten examples