I'd suggest that the time drift was tested directly. 0.5 seconds for most things is a reasonable threshold I'm pretty sure, by the way the OP described the pro

I'd suggest that the time drift was tested directly. 0.5 seconds for most things is a reasonable threshold I'm pretty sure, by the way the OP described the problem that both the service state _and_ the product of the service are being monitored