How do teams handle integration testing reliably as systems scale?

As applications grow with multiple services, APIs, and async workflows, integration testing often becomes flaky or slow—especially in CI pipelines. Shared environments, mocked dependencies drifting from real behavior, and test data setup seem to be common pain points.

How do teams here approach integration testing in practice? Do you rely more on mocks, contract tests, real environments, or recorded interactions? Any lessons learned on keeping integration tests fast and reliable?

I came across a detailed guide that breaks down real-world challenges and practical approaches to
integration testing, which helped clarify a few things.

Would love to hear how others are solving this.