There are various reasons why a composer might write piano music on 3 or even more staves (up to about 5 or 6 on occasions with Sorabji). But I also firmly believe that the majority of music written on three staves could quite conveniently be written on two staves with no real loss of legibility. I suppose composers' tastes vary, and some just seem to like to spread their music out onto three staves. If it's necessary, this should be done in something like a logical fashion, to separate out voices that might otherwise get confused if combined onto one staff. But passages in Book 2 of Debussy's Preludes give me the appearance of quite haphazardly being scattered around the three staves, with no particular pattern to it. I firmly believe that few passages in the entire Book 2 of the Preludes needed to be written on three staves.
You raised J. S. Bach as an example, and suggested that if the purpose of multiple staves was to separate out voices, he would have to write each voice of a fugue on a separate staff. But that would be a bad mistake, and would make reading at the piano extremely difficult. Yes, a principal purpose of extra staves is to separate out voices - but it can't be done automatically every time three or more voices exist - a composer would need to carefully consider when it would help matters, and do it only then. I would tend to do it, for example, only if the voices separated out into two or more groups were rhythmically quite independent from each other. That makes it far easier to read the multiple staves and consolidate their contents into a unified whole in your mind. And to justify that, in my opinion at least, there would have to be considerable and confusing clutter resulting if you forced the music to fit into only two staves. Reading extra staves *can* be more difficult; so the benefit you get from it has to outweigh that.
Scriabin quite often uses three staves - occasionally four. If I had composed that same music, I would have used the normal two staves in most cases. But I don't find his multiple-staff passages difficult to decipher at the piano.
I can cope with three or even four staves with no difficulty at all, so I have never viewed this as even a minor irritant, even when I believed it was not really necessary. In my own composing, despite complex, multi-tiered textures with inner voices and the like, the vast majority of the time I find two staves sufficient; but I don't hesitate to use three, or even four, if I believe it would be better. (I don't think I've yet exceeded four, even briefly.) This policy may partly result from the fact that I tolerate more clutter on a particular staff than some other people do.
As to Rachmaninov's 4 staves in his famous C# minor Prelude - I would consider it a great overreaction to reject playing the piece (or dislike doing so) because of that. While I consider the 4 staves unnecessary and believe two would be entirely sufficient with no increased difficulty in reading, I think the 4-stave passages are extremely easy to read and put together mentally at the piano, all the same. I don't think it is an issue anyone need get hung up on at all - that passage is essentially very simple in its structure - far more so than the majority of passages Rachmaninov notated on only two staves.
It seems to be quite a divisive topic. As for myself, I don't strongly side either way - but slightly prefer two staves, except in those cases where three or more really do make the music easier to read and learn.
I have no rules or procedures for determining when that is. I don't even bother deciding it in published music, because it's the way it is and I have to deal with it.
In composing my own music, I do have to decide staff arrangements, and I just have to evaluate each case on its own merits. I probably decide in favour of two staves 80 or 90 percent of the time, three staves the rest, and maybe four a tiny, tiny percentage of the time. I occasionally use polyrhythmic time signatures, and that can influence a decision to add an extra staff - although two time signatures *can* (at least briefly, or in simpler textures with regular rhythmic structures) be combined on one staff, one stacked on top of the other.
Regards, Michael.