When testing EventGate in kubernetes, I've noticed that each instance will take a long time (many seconds) to load a new schema, and seems to block during this time. I haven't noticed this slow down when testing locally. This might be a problem with EventGate (and AJV?), but it also might be a k8s related problem. I need to investigate more.
In either case I don't think EventGate should block in the way it does when loading a new schema. This could potentially cause request timeouts to a particular pod while the service is loading some schema it hasn't seen before. For schemas that are available locally, we could implement some pre-caching, but this would still be a problem in the future when we start using a remote schema registry to replace EventLogging analytics use cases.