How well it runs.

The console measures the downtime across the period it watched, not across the whole window. A new application has a short history, so the page says how much of the window each figure covers. The console calculates the downtime from the health checks, not from the fault records. The checks are evenly spaced, so the ratio of bad checks is a fair estimate of the ratio of bad time. A fault record starts and ends on a check, so it would round each fault up to the check interval.