In Peter Laackmann and Marcus Janke’s “Uncaging Microchips” talk, at 30'20" they presented an approach I hadn’t heard of for removing epoxy, the most common encapsulant, from a microchip package. By heating ordinary rosin or colophony to its boiling point of 320–360° with a heat gun (which also is not a temperature I knew rosin would withstand without charring) you can dissolve the cured epoxy package in under 20', then clean it with acetone at 40°, though I suspect alcohol might work as well, since what must be removed at that point is mostly colophony. Reportedly it smells terrible and leaves the chip unusable because it loses the bond wires.
They also mention that chloroform, dimethyl formamide, and DCM can swell or dissolve epoxy, so that you can “brush it away”, as they say, although I would rather not be around any of those; and you can use a CNC milling or grinding machine with micron precision; and you can burn it with a laser, especially a 10-micron infrared laser to which the silicon is transparent.
I suspect you could probably burn off the epoxy with a non-thermal oxygen plasma as well; the epoxy’s reaction products with oxygen will be gaseous at room temperature, while the reaction products with the bulk components of the chip — aluminum, silicon, copper, silica, hafnia — either don’t exist or are solid. Maybe a non-thermal steam plasma would also work, because although silane is a gas at room temperature, it’s not very stable. And of course ionization of air generates oxides of nitrogen, which are of course well known as a way to decapsulate epoxy-encapulsated chips; the talk above says you usually need several grams of them. See the note on cold plasma for more.
The rosining process is pretty interesting to me not only for seeing the chip — for example for reverse engineering — but also for the possibility of converting a packaged chip into a WLCSP, since WLCSPs are usually hard to buy, especially in quantity 1. The chip would need to survive its rosining, but I don’t think 360° for 20' is enough to cause substantial dopant diffusion; I think it’s just a question of replacing the broken bond wires.