All services have service-checker software running in production. The checker fetches the service spec, parses it for x-amples, constructs requests and executes them. Currently, for RESTBase this is only run for en.wikipedia.org domain. However lately we're having more endpoints that are exposed on domains other then enwiki.
Definitions for wiktionaries, recommendations for commons and wikidata, AQS on wikimedia to name a few. We need to figure out how to support fetching specs for multiple domains and constructing a full set of tests.