Are you sure its related to Camel's JDBC idempotent repository? If you take it out, or switch it to a memory based etc (though its limited in # of ids it can track). But just want to get it out of the picture.
Also have you setup indexes and whatnot on the JDBC schema/table you use for the JDBC idempotent. So the SQL's still perform?
And if you got any SQL tool that can measure/display the performance then you can see if the SQL gets slower.