by default, vc_ops gets data from vCenter every 5 minutes.
Which means, it will take anywhere between 0 and 5 minutes for the vc_ops to understand that there's a behavioral change in any of the objects being monitored.
In addition to that, after a threshold breach was detected, there's a delay with at (AFAIK) 1 cycle before an alert is trigerred. That's in order to avoid noise.
For the alert to clear, the condition must be "gone" for a few cycles as well.