In the real chess world tournaments play, USCF or FIDE, players are allowed to play against opponents of the same category. In other words a 1500 player is not allowed to play in a tournament against a say a 1200 player.
So far, only Paul and I have tried to choose options other than the 3 given by Erik.
If this thread is intended for survey, the respondence is too small to be valid.
If this thread is intended to seek opinion, it is not right to "force" respondence to choose the 3 options, especially the details of these options have never been clear. There are a lot of issues here. More complicated than the most complicated.
Reading from Erik's post, where he started with quick "introduction" to human strange behaviour (treating rating like money), then followed with how "valid" his understanding of the statistics, I can tell that:
When he started the system, he wanted a perfect rating system. Statistically perfect rating system. What is that? A rating that accurately predicts one's rating relatively to others in the same pool. You MAY want to say that chessdotcom rating system is more accurate than FIDE rating system. If so (actually it is not, the detailed implementation is different), has this solve the problem?
Apparently, many have said that engineers make a lousy management. From marketing perspective, you can see that a lot of consumers want their chessdotcom rating comparable with FIDE rating, even without slight understanding what a rating is! Isn't that ironical?
You can "change" the system from minor change (as changing K-factor or restarting the RD) to major change (from Glicko to ELO with completely different tablebase), all with implications.
How can we make chessdotcom rating "comparable" to FIDE, which has different pool of players?? By using own formula! But of course that's not a good idea. (I don't know which "new formula" Erik was talking about)
Imo, there is nothing wrong with the bullet rating (FIDE doesn't have bullet rating AFAIK). I can see many players only want to play with higher rated players, only want playing with white pieces. If overly high rating is what they want, give them what they want! They are the customers.
I myself don't care with my numbers. But I want a clear way to track "progress". It can sufficiently be done for example by providing "top 9000 players" so that I can see (by printscreen) that today I'm rank X in chessdotcom, Y points below LordNazgul.
I think my slow chess skill is a lot better than my quick chess skill. But my chessdotcom standard is lower than my blitz, and my blitz is lower than my bullet. But it is not because the rating system is not accurate!
I intended to "achieve" 1700 at bullet games and work out through the higher time control. My peak at bullet is 1752. My average opponent's rating is 1672. I think it is normal to have opponent's average rating lower so I decided that I'm done with bullet.
I don't want to hurry playing longer games. What I want is 3/1 or 5/1 because I want to think seriously but don't want to lose on time. But it is difficult (slower) to find opponent with 3/1 or 5/1 so I have to play 3/0 or 5/0.
#2 sounds best to me.