Yep, we use Cypress Cloud. I've found that it's helpful, but not sufficient: most of our flakes are subtle timing things, where you can't tell just from watching a video why something didn't happen. We do also capture the backend logs during the tests -- this is sometimes sufficient, but we don't have a good way to expose the backend logs with the test results, so my engineers often forget that those logs are even available to them. In this context, by integration testing I mean "tests running against a full, working backend, augmented by database fixtures." I've never had much luck with network mocks -- I've found that if the flows you're trying to test are sufficiently complicated, the mocks become tantamount to a reimplementation of the backend, which feels like a bad use of time to me. There's also the issue of API stability: when the API is evolving rapidly, keeping mocks in sync becomes extremely tedious -- the API design at that point is more of an implementation detail than a contract, and asserting around it feels silly.