I like the new look (I think it's more usable than github). In particular I find the "Commits"-tab more intuitive for inspecting code across multiple branches, even though I can't exactly put my finger on the reason. On github I often have to double-check which branch I'm looking at, on GitLab it feels more natural to me.
Basically the only UI-change I would make at this point is on the project-homepage. I'd show those tooltips to the right of the filter-buttons (not above) so they don't cover the other buttons while you hover.
I see nothing questionable in copying a good layout anyway, as long as you don't try to fool the user into thinking you are the product that you are borrowing from. There is no point reinventing an existing wheel unless you can make it significantly rounder.
Agree though, github deserves all the karma we can throw at it.
However, it starts at $5000 per year and the main group amongst people I've heard using GitLab has been companies not willing to pay that $5000 (but who otherwise would probably go with GitHub - that is, they're 'satisficing' rather than choosing GitLab as such).
It's priced more competitively than Github and you get access to the source code.
The file browser sucks, the repo management doesn't make sense (repos live in projects, instead of organization/project or person/project), you can't run git via ssh on port 22, no activity stream (so you need to combine it with Fisheye), etc.
The price of Stash is really the price of Stash + Fisheye. At that point you're only looking at $2500, but still. Unfortunately for us, GitHub issue tracking and wiki suck compared to Jira/Confluence. So we run GH Enterprise + JIRA + Confluence. :)
GitHub is a company that in my opinion can't get enough hacker karma, they have done a brilliant job of innovating time and time again and delivering a brilliant product that makes all our lives better. I can't even bring myself to use open source projects on Google Code, it just isn't worth the time. They always seem to act honorably and do their best to provide an incredible product.
Taking all that hard work, in features AND look and feel and blatantly ripping it off would make me worry about my karma. The authors may just find themselves reincarnated as Cobol programmers in their next life.