If I change a regular unique index on users.email to a case insensitive one by creating this index:
CREATE UNIQUE INDEX user_email_ci_idx ON users ((lower(email)));
Any reason not to drop the prior case sensitive index?
I would guess that postgres will switch to the new index, and the resulting results set would be equivalent in terms of performance.
On a related note, do others use case insensitive unique indexes for users.email?