BUG/MINOR: fix forcing fastinter in "on-error"

health_adjust() should requeue the task after changing its expire timer.

I noticed it on devel servers without load. We have long inter (10 seconds)
and short fasinter (100ms). But according to webserver logs, after a failed
request next check request was called with same 10s interval.

This patch should probably be backported to 1.4 which has the same feature.
1 file changed