They’re seeing bipolar voltages, so I’d be wary of polarized electrolytics there. In the oscillator mentioned below I use 220 nF for fairly low frequencies, but if you need much lower (22 µF), bipolar electrolytics might be a good idea. Anyway, I would not think reversing these caps would do much, since the voltage isn’t polarized. Or at least, it isn’t if it’s oscillating, which it isn’t… Yeah, if it’s not oscillating, and there’s a large enough DC voltage on them, then polarized electrolytics definitely are at risk!
I’D SUGGEST IMMEDIATELY SWAPPING IN THE BIGGEST FILM OR CERAMIC CAP YOU’VE GOT FOR NOW, TO AVOID A LOUD AND POTENTIALLY DAMAGING INCIDENT. Once it’s working you can go back to the electrolytics, though, again, I’d use bipolar ones.
It’s fairly similar to this one
which has some description of how it works. (And I’ve built that one, it does work — simulated it too, in CircuitLab.)
There the feedback and input resistors on the comparator are 220k and 100k, so the comparator flips when the triangle wave reaches about ±6 V. In the schematic you’ve used there’s instead a pot which would allow changing that amplitude, but if it’s centered then the comparator would not flip until the waveform is at about ±10.4 V, i.e. the maximum output of the TL072, and if it’s slightly beyond centered so the feedback resistance is smaller than the input, it seems to me it would not oscillate at all. If so then half the range of the trim pot is useless. And in your build you’ve used two resistors to sit somewhere close to, maybe just the wrong side of, the centered value. Maybe try something like a 20k for the input and 30k for the feedback instead.
Disclaimer: I haven’t had breakfast yet. My ability to think is suspect. The above paragraph may be dead wrong.