A change/improvement to the current skill rating formula that takes rates into account is something I've played with before. One assumption of the skill rating formula is that all songs with the same difficulty are equally difficult. Clearly on an individual level this isn't going to be true, but we have a team of people that makes sure it's approximately true otherwise. This becomes a lot messier when you throw rates into the mix.
My impression from talking to people more experienced than me was that rates cannot be assumed to affect all songs of the same difficulty the same. So for example: if I wanted to make a formula that made a statement like: "Song X, which has difficulty 50, if played at a rate of 1.2x will have a difficulty of 60", then, lacking any other information, that statement would have to be true for ALL songs of difficulty 50, or else the formula would never work. So, unless we do think rates affect all songs of the same difficulty the same, a skill rating formula that takes rates into account would need more information than just the song's difficulty, the rate, and a players score on the song.
There are a variety of data science-y ways in which you could get some data to inform this system, some more viable than others.
I found an old internal staff document that outlined a method of having people judge difficulties of rates, so they could be treated like any other song, but it introduces some difficult technical problems, as well as a massively increased workload for judgement staff.
While I think that the current skill rating formula was a huge step in the right direction (if I do say so myself), and has delivered a lot of enjoyment, it's far from the ideal. I hope that once some other more pressing matters are addressed the staff team will get a chance to push forward in this area.
Incidentally, if anyone here is looking for some way of tracking their improvement over time, but doesn't feel like the current FFR site provides enough control or granularity, consider trying:
FFR Tracker.