First, I'm assuming you don't have a generic plugin running on every record create or associate action.
Have you tried turning on tracing briefly, reproducing the error, and then stopping the trace? The trace file is usually a better way of getting more details about the exact nature of the error.