That's exactly what happens here. There should be some sort of "GPT-3 Summary" of support notes—I've recently had interactions where some problems were documented very clearly by good folks in early notes, but everyone that works on a case afterward tends to only read the most recent 2-3 notes.