There is a failure mode which gets triggered on heavily loaded
systems. A key change is scheduled in 5 seconds to make ZSK2 inactive
and ZSK3 active, but
named takes more than 5 seconds to progress
rndc loadkeys to the query check. At this time the SOA RRset
is already signed by the new ZSK which is not expected to be active
at that point yet.
Split up the checks to test the case where RRsets are signed correctly with the offline KSK (maintained the signature) and the active ZSK. First run, RRsets should be signed with the still active ZSK2, second run RRsets should be signed with the new active ZSK3.
(cherry picked from commit aebb2aaa)
Closes #1413 (closed)