You write tests on your laptop. They do their real job on a build server: a machine that runs your whole suite automatically whenever someone changes the code. Most teams call this CI.
Four things are different there, and each can cause a failure.
The machine is slower, and there is no screen. Build servers are usually smaller than your laptop and run the browser without showing it. Tests that quietly depended on your fast machine start failing. That is not bad luck. The server is finding timing assumptions you did not know you had made.