Improve the way to recover from destination web server failure

When a destination web server fails 9 times, it's disabled for 300secs before it's tried again. We continue to incrementing seconds when fails again.

If a destination fails often or there's no other (enabled) destination, that makes the measurement continue failing.

A simple way to improve this, is to don't "disable" it and just continue trying (but logging warning).

Otherwise we should review the number of times it needs to fail and the number of seconds re-trying.