jQuery is concerned with speed as far as I'm aware; I see benchmarks for it somewhat often, anyway, and I know it's used at a lot of big sites who are also very concerned with speed and load time and therefore has had a lot of iteration and testing with what methods work best.
Of course, if Facebook tested jQuery and found it too slow for their needs, that's all fine and good, but somehow I doubt they have. If there is a specific function that's too slow, there's no reason you can't improve or rewrite that function specifically and keep the rest of the benefits of the library.
I just doubt that Facebook came up with something faster out of the blue. In some cases, it is true that highly-optimized and specific code is necessary to get maximal performance. Those cases are relatively few, I think, and usually involve code much lower level than JavaScript. In the case of JS, the code should be optimized to perform optimally across all of the major JS VMs. This sounds to me like something that jQuery's authors would have familiarity with, as they spend all day working on that.
jQuery is not an abstraction or extra layer of cruft to bog things down, jQuery is a general-purpose library. It's not like an ORM or a dynamic language that adds a whole new abstraction that didn't exist before. Yes, ORMs and/or dynamic languages are slower, but they're slower because they obscure a major functional piece that still has to be performed (mapping onto the appropriate SQL dialect and memory management, respectively) by the computer. In this case, yes, a human could usually make a better choice than a computer if that human was adequately informed.
jQuery is not like that. jQuery is not an abstraction, it doesn't abstract away any major task that the computer has assumed for you. It's just a fast library of commonly-used functions and building blocks. It's frequently tested against new browser releases for compatibility and speed. It's frequently optimized and improved upon. There is a huge community out there on tap for support and there are a lot of plugins for it. It's open-source and easy to modify if your case requires it. There are lots of other people from other places that are making important and great changes to that base at no cost whatsoever to you or your company, changes that you can apply just by installing the new version (in backward-compatible releases, naturally) and magically enjoying this speed boost that some other guy provided your website.
I don't see any reason not to use a good open-source JavaScript library. We've used jQuery in this example, but I'm sure it's applicable to many others. I don't understand any business reason why Facebook would want to write something completely from scratch when there are these many frameworks already available. The only reason that I can see why this would be implemented in this way is either developer indulgence or developer ignorance.