nitpicky culture, we maybe have that problem with Openstack where there are thousands of developers, but for our projects in SQLAlchemy we're a team of about five people and I'm more or less a BDFL type of role, to the degree that we are nitpicky about things it only prevents much bigger problems from happening later, if a review has little things that are bugging me I'll just fix them myself and push a new change up rather than bothering them with it, also something you can't usually do with pull requests.
I think Wikimedia struggles a lot with code review culture in general. Different people have conflicting ideas about what good code looks like. It used to be very nitpicky (i've had code rejected in the past for using (php's) intval() instead of casting to int. I've also had code rejected for casting to int instead intval().) But that's improved quite a bit with better precommit lint tools. The length of the feedback cycle is very long and sometimes feels like its mostly about who you know (e.g. the last patch i submitted i did on sept 9 for a decently serious bug. First actionable feedback (which was relatively minor things of the form use a constant named ONE_MINUTE instead of 60) was on oct 15. Thats kind of a long time to wait for code review imo). Anyways, its just not fun to contribute when code review is so unpredictable and long.
Hmm. Guess i got off on a bit of a tagent there. I do think gerrit has some usability issues, but i think that's hardly the main problem.
Can strongly recommend to remove "-1 code review" and require all comments to be resolved instead. Accomplishes the same goal why being more positive.
I'm pretty sure most of the anti-gerrit sentiment at wikimedia was about gerrit as a code review tool.
My personal experience with it (as a mediawiki developer) is gerrit has a lot of UI bugs (although it has gotten better). I also suspect it encourages a code review culture that is overly nitpicky and risk averse (but perhaps that is just cultural forces at wikimedia)