@18: I'll second that, with the caveat that you can't change someone else's mind with snark if their mind is made up, but that undecided bystanders are often swayed by wit and humor. If one side is dead serious and blustering while the other is light and witty, that's good evidence that the dead serious side is acting defensively because they don't have reality on their side. Of course, bystanders are turned off by outright cruelty, or if the witty side's criticisms just don't make any sense. And the repercussions of this effect are far from uniformly positive....

@19: "Meta" is Greek for "beyond", and when used in English as a stand-alone noun instead of an adjectival prefix it clearly means the postmodern lit-crit sense "[analysis] beyond [the surface meaning]": meta level 0 is directly making a statement/argument/rebuttal, meta level 1 is questioning the motives behind a meta level 0 statement or otherwise addressing its form or context (e.g. someone is clearly making a dishonest argument because of a conflict of interest, so you call them out on it), meta level 2 is addressing the form/context/motives of a meta level 1 statement (e.g. you were making an earnest argument when they questioned your own motives, so you retaliate by accusing them of dodging the question because they don't like the answer), and in general a meta level n+1 statement is addressing a meta level n statement's form/context/motives instead of its substance. When "meta" is used in these senses, communication of meaning has been achieved (the entire point of language), so the usage is perfectly cromulent, proscriptions against verbing and nouning things be damned.

My own use in @17 wasn't strictly hewing to this sense, of course; it was tweaking it for the sake of a math/comp-sci modulus 2 wraparound joke. It's a long-standing norm that jokes get to abuse language for Rule of Funny. To state the point seriously, a really good snark attack will demolish a foundational premise of the bad argument (meta level 0), yet that will be veiled behind a witty attack on the form/context/motives of the bad argument (meta level 1). If you got that from the joke and didn't need it explained, then I succeeded at communication and there was no need to criticize the joke's form in the first place.

It's worth noting that a good amount of the "meta" abuse out there (perhaps even "[x] is so meta" for "[x] frequently uses self-reference") originated from or was influenced by comp-sci geeks: 60's MIT hacker culture turned "meta" abuse into a sport/joke, and it spread from there into broader geek culture and thence to the Internet. For instance, while "metadata" a.k.a. "data about data" mostly hews to the postmodern sense of "meta-" (it answers questions like "who created this data", "where did this data come from"), "metalanguages" instead are "languages for describing grammars" -- NOT languages for expressing "who invented this grammar", "where does this grammar fit in the Chomsky hierarchy", etc. but merely for compactly expressing which strings of symbols are permitted by the language's rules. EBNF such as «mood ::= "happy"|"sad"|"angry"; stmt ::= "I am a " mood " cow."» is a metalanguage in the comp-sci sense, but there's absolutely nothing postmodern about EBNF because it doesn't even acknowledge the existence of semantics in the base language whose grammar is being depicted (the tokens "happy" and "sad" are 100% interchangeable). The "meta-" prefix here has been divorced from its lit-crit sense and generalized to "any recursive structure" or "any reference between two things of the same type".