@caughtinarut You make some good arguments which I appreciate but your points are not quite right. First, rating systems that measure chess performance are actually objective. Objective just means, well let me quote, "(of a person or their judgment) not influenced by personal feelings or opinions in considering and representing facts.". Glicko only considers win/loss/draw, opponent's rating and some other technical numbers. It's impervious to personal feelings or opinions. What you may have been thinking of is the difference between something that has an absolute value vs something that has a relative value. And while chess ability certainly is relative, that in no way prevents us from obtaining objective measurments regarding it. Speed for example is also relative, but we use instuments to measure speed all the time.
The composition of the player pool is an interesting and somewhat complicated matter but it can be handled very effectively. First your example about large numbers of low ability or high ability players all joining at once is very hypothetical because the size of the Lichess player pool is sufficiently large that the likelyhood of overall lichess ratings being thrown off in this manner is very low. It would require something truly weird happening. Furthermore it's well established that the composition of almost all player pools including Lichess and FIDE follows a normal distrubution/ bell shaped curve.
What could pose a problem though is that the overall ability level is higher in the FIDE player pool. However the solution to this is not nearly as drastic as the one you suggested. You don't need all Lichess players to have a FIDE rating you only need a healthy sample. If we just have a few players with FIDE ratings that also play here with their best efforts, then there are enough data points to see how to adjust Lichess ratings to closely match FIDE ELO. You could do periodic calibrations to fine tune it better but it wouldn't be that important as the overall composition of both is relatively stable.
I'd like to hear you tell me caughtinarut, or anyone, why this wouldn't work?