As cited by Z (a/k/a Olivier Fouquet), a commenter at Crooked Timber:
In fact, almost all non trivial results of […] can be put in the following tripartite classification:
(a) Results for which a (sound) reference is given, but of which no proof can be found in the given reference (e.g. loc. cit., Prop. 4.2.2(i)).
(b) Results for which “a proof will appear elsewhere”, and for which no proof has appeared anywhere (not even in a preprint, twenty months after the publication of […]). (See e.g. Theorem 3.2.1 loc. cit.)
(c) Results for which a reference is given, that does not exist. Indeed, many results are said to be proved in […]. This preprint does not exist, either in paper or electronic form, under this name or any other, and for obvious reasons will never do.
It’s well known that mathematicians love elegant proofs; they clearly love elegant take-downs as well.