It's very good at telling you where your program is spending its time. But if what's running is CPU bound code, it can be tough to get a full picture of that. So even across things like pandas or NumPy, potentially, it could misrepresent how much time you're spending there. On the other hand, it was never going to interfere with that raster or C, but it's also not going to report inside that. It's one of these Heisenberg. sort of things by observing that you make a change.