A study by GMAT say that last few questions does not change your percentile that much.
Please read that full study. This is true if you are scoring at a "medium" level - or in the 500 to 600 range. It is NOT true if you are going for a much higher score!
Missing one question anywhere won't hurt your score all that much. But the GMAT is a "where you end is what you get" test, so if you increase the overall rate of missed questions in the last quarter of the test, your score is going to go down.
To answer your question: first, do not wait until question 31 to start dealing with the fact that you're 5-ish minutes behind.
Your best strategy is to identify the hardest questions for you
as you see them througout the exam - and make your guesses then. You're probably going to get them wrong anyway (they're hard!), so get them wrong fast.
Read this right now, then come back here:
https://www.manhattanprep.com/gmat/blog ... -the-gmat/Okay, so how do you know which are the hardest questions for you? Data and analysis. Look at your recent (timed) performance. Look for things that take way too much time (>2m for SC, >3m for CR, >2.5m for RC). Think about two categories here: wrong ones, and ones you got right but in seriously too much time. If you get a CR in 3:01, okay, that's not too terrible. But if you get a CR right in 4:26, that's terrible, because it just cost you
at least one other question elsewhere, and probably more than one. (Mental fatigue is a factor on this test too, and getting one CR right in 4:26 is way more mentally fatiguing than getting two questions right in 4:26. Something's seriously wrong or it wouldn't take you 4:26 in the first place.)
Start looking for patterns. You might realize that you hate CR Evaluate, or SC where the entire sentence is underlined, or whatever. Can you generally do part of that type of problem in reasonable time, narrowing down answers effectively, but not the whole thing down to one answer? If so, do that (in reasonable time), then pick from whatever's left and move on. If even that's a struggle, just guess immediately in future when presented with that set-up.
Note: you can't be too broad on this, or you'll be guessing on too many problems. For instance, you can't say that you dislike all RC detail questions - that covers about 3/4 of RC questions.
You can already do this in hindsight: you have looked at test results before and thought, wow, that problem was terrible. I probably should've just moved on. The trick is to learn how to spot that in the first 30 seconds, so that you really can move on. You can do it! You just have to study how to do this, just like you've been studying everything else.