Eric mentions that vblank handling improvements were accomplished last week along with addressing VT switching bugs. The next high priority item to be addressed though is with batch-buffers failing due to large sizes. When not using GEM this issue has been addressed, but right now if you're using this kernel memory manager you can run into issues that result in an error message and no rendering when a batch-buffer is too large and can't be allocated.
The Intel developers have also changed about their procedures in which they submit patches (back to attaching patches and not Git trees to pull from), in order to foster more code review.
Eric's blog post about Intel graphics on Linux can be read here.