Mock dependencies for tests

I need test doubles for a NetworkClient dependency used by SyncService: provide a mock that can simulate successful responses, timeouts, and HTTP 500 errors, and a fake that records requests and returns canned responses synchronously for deterministic unit tests. Show how to inject each into SyncService in tests and assert behavior on retry and backoff logic.

Before I add these mocks to the test suite, make them easy to review: put the failure modes (timeout, server error, malformed payload) up top, make the canned responses findable, flag where behavior diverges between the mock and the real client (threading, latencies), and suggest one small change to make the mock reusable across other services.

I built a mock NetworkClient that returns an error code to simulate a timeout, but SyncService's retry logic does not run during the test. I need to say what I implemented, which tests fail, who expects retries, and what I fear (the mock not matching the client's failure semantics or missing async scheduling). What's the likely cause and the best fix so tests exercise the retry logic correctly?

Across the codebase, every team writes their own ad hoc mocks for network and storage clients, and they slowly drift from the real clients' behavior, causing false confidence in tests. Which habit should I change to prevent this: recommend a shared mock/fake library with canonical failure modes, a periodic sync test against a real client, and one enforcement rule in reviews to keep mocks honest.