..
Without further experiments I can't say with certainty what really helped, but it appears that the key to success is high temperature and reflux.
..
The colophony contains acids. Under ambient temp the acids are not active. With higher temp they activate.
Therefore colophony has been used as the soldering flux since ever (there are many varieties of the colophony too)..
But you can't go too hot either, or you end up with carbonized mess like on my first attempt 4 years ago:
Further experiments will be needed to find the best temperature.
Here Sacodepatatas says that it should only take a few minutes, so maybe I still didn't heat it enough. Experiments will also be needed to see if this works on anything other than AliExpress junk, because the chip I decapped was a Chinese MC34063.
It's not clear what the exact mechanism is. Maybe there is a reaction with resin acids and temperature speeds it up, maybe at 380°C epoxy pyrolyzes by itself and colophony only acts as a solvent for the resulting products. Maybe the resin acids turn into something else, more aggressive towards epoxy.
One thing I can say is that boiling for many minutes at 340°C does nothing, but at 380°C or so something starts to happen, the solution turns brown-green and loses transparency, and the chip disappears.