Glent should ignore queries that have search syntax in them (dog*, !dog, -dog, dog AND cat, intitle:dog, etc.), but rather than try to have Glent figure out what is or is not search syntax, Cirrus should share what it knows.
Once the info from Cirrus is available, Glent should discard M1 queries and suggestions that have search syntax in them. (It might not be terrible to do this for M0 and M2, but M1 really needs it!)
We probably do not have to re-process historical data (it should all be re-processed each week), but we should double check that that appears to be the case once this is up and running.