Yeah that’s a pretty crappy situation for a new contributor. You basically have to make some educated guesses. Do more tests fail on average after your patch than before? Are the fails in tests that you added or in tests that look related (test with backup in the name for example)?
If the fails doesn’t seem related to your changes, you pretty much leave it on whoever might commit your code to use their deeper experience or luckier dev environment to get your changes in regards to unrelated tests. That pretty much dictates you want to run the tests a fair amount to help you gauge. To do that reasonably, you need to crank up the parallel. Almost all of the longer tests do very little actual work - so you can run a lot of them together as long as you have the RAM for it. On any decent modern hardware, you ought to be able to run them in at most 40 some minutes in most cases. With good hardware and the right parralel test settings, you should be able to do more like 10-20 min.
