Our scollector and collect duration timings are probably being effected by using time.Time which is not monotonic and can be effected by changes to the system time (manual or automated by NTP). Not a huge deal, but likely causes incorrect data in the duration metrics.

Access to a monotonic clock is slated for go 1.8, but there are also packages like goarista and monotime. We could also use it as an opportunity to switch these metrics to use atomic operations instead of locks.