On managing socket's error queue

Is there a common design pattern to manage 'sk_error_queue' of 'struct sock'?
I've found that some socket destructors (say, 'iucv_sock_destruct()') do
'skb_queue_purge(&sk->sk_error_queue)' but (most of the?) others do not,
and it would be very helpful to know why it is so.

This question comes from https://syzkaller.appspot.com/bug?extid=7ff4013eabad1407b70a
where the trivial patch (hopefully) fixes a memory leak.

Dmitry