For what it's worth, the citext_pattern_ops option seems to provide case-blind = searches, which text_pattern_ops does not.
That is correct, but the default operator for citext type already provides case-blind =. The only improvement on that that citext_pattern_ops could hope to achieve is fast case-blind prefix matching, which it fails to do.
Note that you can get case-blind LIKE matching use ILIKE, and can accelerate it with a pg_trgm index. However, if the only type of matching you want to accelerate is prefix matching (% only at the end of the pattner), then pg_trgm will be much less efficient than a fully-functioning citext_pattern_ops would have been had it done what its name implies. Still, it might be better than the other choices you currently have.
