Adding a little bit more information:
When the issue cropped up again, I tried running
"unbound-control dump_requestlist" to inspect the list, expecting it to
of requests that are stuck for some reason. Instead I found that it was
completely empty. Yet incoming queries were constantly being rejected.
Then I tried "unbound-control flush_requestlist" to see if it had any
and sure enough, it resolves the issue, even though there doesn't appear
anything to flush.
When I get a spare moment, I'll try to add some more logging in that
part of the
code, see if I can specifically determine what condition is failing that
causing the queries to be dropped.