If the issue happens that infrequently then logging the error with as much context you can provide is the most sensible thing you can do.

The best fix is correcting the issue rather than dropping data on the floor and retrying. The worrying thing is that if you can get obvious corruption of messages you may also be able to get subtle corruption that looks OK to your tests on the receiving end, but is wrong. If you have control over both ends of the link you could change the message format to include error checking so that you have reasonable confidence that your data is reliable.

This is not a problem that we can sensibly help with without a lot more context. In the worst case this sort of problem can kill people and in some jurisdictions we could be held libel as a result of giving you unfortunate advice.