I made an earlier comment at the wrong level - both technically - to the wrong part of the discussion, and also from the purely selfish perspective that the change in Glibc behaviour could break existing systems - potentially costing me my cash.
An earlier suggestion was to replace the change with a API violation detector that causes programs to crash rather than corrupt their state.
This is better than a silent corruption - but still antisocial. A program that used to work now fails. Not everyone can fix the cause of the crash. Not all software is unimportant to the user. It is a bit like suggesting that when you see a j-walker you should just mug the person next to them, as a deterrent for future jay-walkers.
I think it is fine that testing releases use the crash-bad-behaving applications change.
But the system released to users should provide the previously working Glibc.