Can somebody tell me any details about this "drift compensation"? How does it work? Does it periodically measure the same test pattern (native display black and/or white and/or primaries?) and calculates compensation?
Yes. See the documentation <http://www.argyllcms.com/doc/dispcal.html#I>.
How do I tell if I may need it or absolutely not?
If you get problems that are due to the instrument black calibration drift, or don't want to spend the time acclimatizing the instrument to the display conditions, then they might help. If you don't have these problems (or don't understand the problem or what the option attempts to do), then don't use them. [The point is that I can't say if these options sometimes help, always help, or sometimes cause problems themselves, or always cause problems themselves, because I can't test calibration and profiling for every possible purpose, with every possible instrument, with every possible display, in every possible measurement situation. All I can do is put it out there after a limited amount of testing, in the hope that it might be useful, and see what people report.] Graeme Gill.