Today I decapped more stuff with colophony: TL062, REF02, LM317L and BC560, so it looks like this should work on every epoxy chip. For some of them I used colophony from a different source, no problem with that either.
Heat appears to be the most important factor, with more flame I managed to reach 390~410°C and things were going faster. One TO92 package was almost completely dissolved after 20 minutes. Even if the chip doesn't fully disappear, remaining epoxy appears to be soaked with colophony and becomes quite soft, so it can be removed manually. By the way, the same happens to heat shrink tubes working in atmosphere of hot colohpony vapor - they soak up, soften and fall apart. It's possible that many plastics can be destroyed this way.
I have not succeeded in getting a fully preserved TO92 package. It looks like intense bubbling separates the die every time. All chips I decapped so far were either blown into pieces or still had some epoxy left, holding them together.
In fact, I wonder if this is even a chemical process at all, or purely mechanical - colophony penetrates into epoxy, vaporizes inside it and blows tiny fragments away. No idea...
My test tube is 8cm long and at 400°C it already works hard, condensing and returning a drop of colophony about every second, but that's not enough and plenty of fumes escape. I plan to buy longer tubes and see if it gets better. I have no idea how people are able to run this process in smaller containers without significant reflux. Maybe apply even more heat and pray that the chip is gone before the colophony?
At this temperature colophony starts to carbonize, leaving ugly black residue in the glass. In the last run today, I plugged the test tube with a ball of aluminium foil in hope of reducing evaporation loss. It didn't help much, pressure inside the tube pushed vapor outside, but it appears to have eliminated the carbonization problem - the test tube was clean, and even some crap left from the previous run got washed away. It looks like carbonization only occurs in presence of oxygen and plugging is a good idea.