PG Bug reporting form <noreply@postgresql.org> writes:
> pg_terminate_backend failed to terminate active queries on a materialized
> view until an "idle in transaction" backend tied to the same view was
> terminated.
Can't do much with this report if you don't give a full recipe for
reproducing it.
I guessed that the problem might be "the termination isn't effective
while the session is blocked on a lock", but in an experiment here
that worked for me. So I have no idea what was happening in
your case.
regards, tom lane