That's a good question (and one that I have kicked around mentally myself). Of the three, I would probably let the 85mm go first. But here's the thing: I also have the EOS M, and I find the 85 a really compelling option on it, too. Because it is reasonably inexpensive, I'll probably hang onto it. I do grab the 100L for some portrait and event work over the 135L for several reasons. One is focal length (sometimes 100mm is what you want). The second is stabilization. The third is weather sealing (neither the 85 or 135L have it). One final reason I might grab it is for tight framing of faces where I want to get closer than what the 135L allows.
I do find focus on my 85mm f/1.8 perhaps the most inconsistent of all my lenses despite very careful AFMA. It is one of the oldest lenses in my kit (in terms of design...and actual age) and I just don't think it is quite as good. I am accustomed to a very, very high success rate of perfectly sharp images in my work, and I find that I am surprised by a slight ooF hit from my 85 most often. It's not bad at all, but just slightly worse.
Thanks for that answer. I also happen to have the EOS M, but have used it mainly for a compact travel camera with the pancake, so I've never tried out the 85. I'll have to give that a go.
I've often contemplated what I would do if I was to start over building a set of primes. My strong personal preference is 50mm over 35mm, so that's my starting point. From a focal length standpoint, a doubling approach seems logical, so a 24 - 50 - 100 - 200 collection would be very evenly spaced and would provide the added benefit of IS and macro capability at 100. My hangup with that scenario is the 135, though; it's just too good to leave out!
I know the traditional 35 - 85 - 135 collection handles that issue better (although you would still have closely related lengths if you decided to add the 100 for macro), but changing from 50 to 35 as my primary normal focal length just doesn't work for me.