I've been mapping for a while now. I have some issues with the way OSM currently works.
The current tagging/import approval process is silly. To get a new tag approved formally, you are supposed to post about it on tagging-l. Then you are supposed to write a wiki page, then have comments on the wiki page, then have an RfC on the wiki talk page, then announce the RfC on the mailing list. If someone says "yeah, I'm not wild about this idea", that's pretty much a blocker for it happening at all. The alternative to this process is to simply start putting tags on things when you edit them and create an informal convention. Because the former process is so ludicrously painful and slow, it is much easier to tell people who you meet who say "I'd really like to start putting X on OSM" to just go ahead and do it quietly without telling anyone, and then it probably won't be a problem.
The wiki doesn't match reality. The RfC process is so ludicrously over-complicated partly because nobody is sure whether the wiki exists to document existing behaviour or to specify what ought to be done. That is, nobody is sure if the wiki is descriptive or normative. It has thus ended up being both and neither.
There doesn't seem to be a process for ever tidying the wiki, and no feeling that we should ever bring discussions to an actual consensus. Over on Wikipedia, for all its faults, admins do at some point close discussions and say "here's the rough consensus: X". On OSM, the discussions just rumble on forever.
The culmination of this slow and silly process for approving proposed tags or imports is that now when friends come to me and say "I've got this really awesome idea for using OSM and we could contribute data back", I now say "yeah, good luck with that: the community process is so slow and bureaucratic, it often isn't worth bothering with". That's toxic. It's absolutely ghastly that smart, informed programmer types come to me and say "we've potentially got some amazing data we could contribute to OSM" and my answer is "that sounds awesome but the process is so broken that you shouldn't bother doing it officially, just do it quietly and unofficially".
There's some simple things we could do to fix this.
Close most mailing lists. Mailing lists suck. Over at microformats.org we eliminated all our mailing lists and just use a wiki instead. See wiki is better than email.
Get rid of most of the RfC process. It doesn't work.
Switch to a default of "let's allow things to happen, then fix them when they break" rather than the current "let's have long, pointless mailing list discussions and then say no".
Develop a process that is based around evidence collection. When I proposed the dress code feature, I provided comprehensive examples of different types of dress code that people were actually publishing on the web. The point is that when deciding on a new tag or feature proposal, we should be guided by the evidence of what people are already trying to express in other settings. The alternative process to this is called "making shit up out of whole cloth". It has a proud and noble tradition in standards-making. Alas, that tradition is also accompanied by a history of complete failure.
If we want enthusiastic contributors to work on making OSM fill all sorts of interesting and unpredictable new use cases, we should welcome those people in rather than keep around processes that are slow, bureaucratic and alienating.