It had definite rules, clear scoring, and added guidelines to ensure clarity, while leaving room for creativity.
I think the comments on the question itself make clear that this is a rose-tinted view. If nothing else, the fact that the only person to answer had no idea what they scored should send you an unmistakable message that the scoring is not clear.
I got ... a sarcastic rules lawyer ...
I don't see any sarcasm. Rules lawyering is par for the course on this site: that's part of the reason for writing a solid spec, and partly why we have a sandbox.
... a pessimistic moderator ...
I think that you and whoever that is are the only people who know what you're talking about here.
... and a bunch of close votes.
Close votes aren't a kiss of death. Fix the problems with the question (if possible) and you can try to get some reopen votes. Although I must say that in my opinion there's one unfixable problem: even if you specify the scoring properly and tighten the restrictions on what constitutes a suitable speech, and even if you rule out things like Whitespace answers which break your scoring system, your current definition of "obfuscation" is so permissive that it's not an interesting challenge.
Does this community want anything other than code golf?
I don't speak for the community, and judging by the number of upvotes some questions which I think are dreadful get my opinions on what makes a good question are not shared by everyone. But what I want from questions are
- Clear statement of task. I do requirements gathering at work, and I understand that non-programmers think in terms of their primary use cases and not in terms of corner cases, but if you're participating in a programming-for-fun site than that doesn't apply to you. I expect people who submit a question to this site to be able to look at a spec they've just written and spot ambiguities. It's fine for a question to have some flexibility, but that flexibility should be explained (e.g. "You can take input from stdin, via command line arguments, or via a GUI"; "Write a program or a function that ..."; etc. rather than "Write some code which takes input and ...").
- Clear statement of winning criteria. Ideally I should be able to write a computer program which implements the criteria exactly. (Note that I am anti-popularity-contest).
- Winning criteria which actually distinguish entries. E.g. a challenge which is judged on big-O complexity and which has a standard solution that is provably optimal is going to result in an n-way tie, which is pretty pointless.
- An interesting problem. This doesn't mean that it has to take me a couple of hours of background reading to understand how to tackle it, but if there is a correct solution in less than 10 characters (and you'd be surprised how often people submit questions which have 0- or 1-character solutions) then where's the fun? (On a related note: near-duplicates are unlikely to meet this criterion. If I were you, I would have held off on posting Mirror quine for a couple of months because two quine questions in one day is a bit light on variety).
I'm not against non-code-golf problems. However, I can't think of very many good ones that have been asked. Asking a good question is hard work; asking a good question with a good non-code-golf winning criterion is even harder.