Actually I suspect there are very few CAA-based issuance failures (certainly compared to the number of attempts done). In order to be a CAA failure I believe the challenges would need to have already succeeded, so you're talking about cases where (1) There's a CAA record at all, (2) the CAA record is one preventing Let's Encrypt, (3) there's an ACME client pointed to Let's Encrypt running somewhere that can fulfill challenges and yet is trying to renew unsuccessfully. (And to really be a recurring problem, add (4) nobody is noticing the failures thus allowing it to happen regularly.)
I suspect that in most cases where there's a failure preventing issuance that nobody seems to be noticing or correcting, the challenge would be failing as well (like, the domain name was moved to new hosting but nobody's turned off the old server yet). There was a recent post saying that 80% of HTTP-01 challenges fail. I think that case is much more likely than somebody changing their CAA record to no longer allow Let's Encrypt, but still have a client running that's trying. I'm sure the number of regular persistent CAA failures isn't zero, but limiting iodef notifications to, say, only happen once, or once per few months, would probably mean that they'd be sending iodef email significantly less often than the renewal notices they already send.
I don't have any real data though, and my guesses about the problems at the scale of this project have been proven wrong in the past, so I may yet be wrong once again. 