The three primary problems I have with GNU Arch are:
- The user interface is nearly unusable. Without zsh tab completions, it is unusable.
- Arch is very slow over a network.
- Arch is very slow not over a network.
However, Arch is only VCS that makes branching and merging so natural that I do it regularly. As a result, I develop software and even documents in a way that is fundamentally differently (and better) than the way I have in the past.
Arch is the worst kind of good software; or perhaps the best kind of bad software. The more you use it, the more you hate it; unfortunately, the more you use it, the more you hate all the alternatives more.
Using arch is not a type of masochism, but sometimes it reminds me of masochism a tiny bit.



Responses to This Post
http://codeville.org/
Disclamer. I have the devloper at hand so when somthing brakes, rare, or is missing I can just beet him with a stick untill he fixes it.
The lack of a distributed model is a killer for subversion. It is very annoying in a large number of ways. Consequently, merging things back tends to be much much more difficult and there is no support for things like history sensenative merging, cherry picking, etc. etc. Subversion is a different paradigm so it should not necessarily do all of these things.
If you have never used and really understood a distributed system, I suggest you try it. I've found that understanding distributed version control is like building up a vocabulary with which you can describe and solve problems. Without the vocabulary, you can't really imagine why it would be useful. I didn't until I was basically forced to learn Arch.
There is a system called SVK which is a full distributed system based on top of SVN which much of the same interface. I am extremely interested in that.
svk is wonderful, though overkill for the small projects I'm working on. It is a little more complicated than darcs, but it seems impressively powerful.
Good luck on your search!