Allan had been making very fast progress on TextMate for the first several years, and the community was full of excited “early adopter” types, who were very chatty and supportive, and were actively engaged in improving the TextMate language “bundles”, trying out new features, etc. I found the ##textmate IRC channel at the time to be the best place to get technical help with pretty much any programming language.
Then at some point Allan decided that he had made some suboptimal design choices in pretty much every component of TextMate 1, and wanted to improve those with new designs. But he thought it would take more work to incrementally swap new parts in that were compatible with all of the other stuff he wanted to eventually replace, so he started in on what was a substantial rewrite of everything.
That (a) took away his incremental improvement of TextMate 1, (b) took his time away from being as responsive to user questions/ideas, (c) caused bundle authors to slow down on improvements to TM1 bundles while they waited to see what new features TM 2 had in store.
At the same time, the creation of an "insider" IRC channel took away some of the activity from the main IRC chat, the migration from one big SVN repository to a bunch of separate per-language Git repositories damped a lot of the bundle development activity as people were no longer exposed to every change to every bundle, so the bundles didn’t cross-pollinate as much.
Many of the core people in the community gradually drifted away, and weren’t being replaced by new people, from some combination of existing bundles being good enough, and less overall buzz.
As Allan was getting less feedback about the code he was working on, and less interaction overall from users, he became less motivated. As the TextMate 2 project dragged past its original timeline, both Allan and others in the community started to get discouraged. I would speculate he started to feel like more of the work was a chore rather than a joyful adventure.
There are many great TextMate 2 features which have still not been properly explored by bundle authors all these years later. It could use more screencasts, more technical documentation, more people testing zany things and making feature requests...
For my money TextMate is still the best editor available, from a purely technical perspective. The experience of customizing one’s own editor environment has less friction than pretty much anything else out there. Almost completely nontechnical people can do amazing things with it. But the community isn’t the same as it was.
* * *
Aside: Sublime Text was an unauthorized (and for at least the first few years half-baked) rip-off which borrowed TextMate bundles wholesale without contributing anything back whatsoever. In early versions [and maybe still?] people couldn’t actually edit all of the bundle items in Sublime, so its users were disempowered from customizing their own experience, arguably the main point of TextMate’s design. In the early days many of them came to complain at us about TextMate bundle items which didn’t work in Sublime. In my opinion the Sublime author profoundly missed the point of what TextMate was about and why it was special, and viewed it as nothing more than a convenient pile of existing work he could exploit to kickstart his own app.