Comment 2 for bug 575774

You should keep an eye on whether this messages get repeated more than the three times and also whether the increment at some point is more substantial. Reading through the upstream report, the problematic cases seem to have been the ones where the increment went into the 100ms range.
If this stays at around three times and you see no other ill effect, this could just be indications of the system adapting to hardware restrictions. On the other hand there are some hints that increasing the retries might be a solution. Though we should see what upstream thinks of it.