As I see it, audits have two primary roles.
The first of these is to educate new reviewers on how to properly review. If the reviewer fails the audit, they get a message explaining what to differently next time, and if they pass they get a message encouraging them that they are doing the right thing. This is good.
The second role audits play is to catch out sloppy reviewing, whether by a new or a seasoned reviewer. In this case, education isn't so important - what's important is letting the user know when they've made a mistake, and keeping track of users that are consistently making mistakes (e.g. robo-reviewers).
Note that the "audit passed" messages are only really useful in the first case - once a user has been educated on how to review, we only really care if they still keep getting it wrong. For a reviewer who has learned how to review properly and is doing a good job, the audit passed messages only serve to slow down reviewing by adding extra stuff they need to wade through.
I don't propose reducing the number of audits given to "educated" reviewers - they're still as capable of making honest mistakes and/or losing attention as a new reviewer is. Instead, I propose that educated reviewers shouldn't be given audit passed messages - they should just proceed to the next review task, the same as if the audit were actually an ordinary review. They should definitely still get failed audit messages, however.
As for what constitutes an "educated" reviewer, I think five consecutive passed audits should be sufficient. An educated reviewer who fails an audit should be reset to being a "new" reviewer, at which point they would need five consecutive passed audits to again be considered an educated reviewer. This should be on a per-queue basis - being educated in one review type should not imply being educated in a different queue.
EDIT: re-reading at the question, I'm not really sure if this is basically what the OP is proposing or not. This seems to be the solution the OP has hinted at in the question title, but the final paragraph of the question suggests he's actually interested in reducing or eliminating audits for educated users (which I'd somewhat oppose). Either way, I'll leave this here in the hope that it proves useful somehow.
: Possibly some reduction is in order, but that's not the point of the proposal. Even if it were appropriate, I'd suggest that dropping the audit rate by more than 50% or so would greatly increase the risk of honest mistakes slipping through between audits.