How do you test that two services still work together without spinning up the whole system?

Use contract testing. The consumer defines the requests it makes and the response shape it depends on, that expectation is published as a contract, and the provider verifies it inside its own pipeline. Both sides then test independently and fast, and a breaking change fails the provider's build rather than surfacing in a shared environment days later.

Full end to end environments do not scale past a handful of services. They are slow, they need every team's latest build to be healthy at once, and when something goes red the first question is always whether it is a real defect or the environment. Contract testing sidesteps that. The consumer writes tests against a local stub of the provider, and those interactions get recorded as a contract: for this request, I depend on these fields with these types. That contract is published, and the provider's own pipeline replays it against the real provider implementation. If someone renames a field or changes a type, their build fails, in their repo, minutes after the change, with a message naming which consumer breaks. That is the whole value: the feedback lands where the change was made. The nuance worth stating is that a contract only covers what the consumer actually uses, not the provider's full surface, so it does not replace the provider's own functional tests, and it verifies compatibility rather than business correctness. I still keep a small end to end suite for the few journeys that genuinely need proving across services.