Now that I finally got around trying to repro this, there are two observations: - I cannot reproduce any network hangs. Can anybody confirm these hangs still occur with kernel 4.2.x? If so, as detailed as possible a description under what kind of load they occur (ideally also enumerating cases where they don't occur) would be necessary. Without being able to repro the issue I don't see ways to reasonably debug this myself. - I can trigger the "Too many frags" messages, but having looked through old logs I can chase them back to at least summer 2013, i.e. I'm not convinced that issue is really related to the one here. Still I am looking into how to eliminate those.