Over time my network had slipped a few minutes, but everything was synchronised, so there were no authentication problems. However it was becoming more annoying (especially when email replies appeared to land a minute or two before the original was sent!).
So eventually I was persuaded to put time in to fix it, in the past I just fixed the clock on the master DC, and all was well. However this time…
Everything I did had no effect, DC’s were changed to point to external NTP sources, but even with manual time changes at the command line they snapped back to the wrong time almost instantly.
Then something twigged. My master time source DC was one of the last servers I moved off the old VMWare cluster into a new Hyper-V setup. And of course Time Synchronisation from host to guest was on by default. As soon as the DC changed time, Hyper-V tools snapped it right back. And as the host took it’s time from AD, it was always out. Before, when on vCentre, the host was synchronised to NTP servers, and the time sync worked.
This morning I unchecked time sync as below, reset the DC’s clock, and all was well. I should have thought of this much earlier in the diagnostics. But I guess having moved away from infrastructure and back to dev, my mindset has changed a bit! These days I’m just a consumer of the network infrastructure (although I do have to be the admin too!). I suppose I could have just changed the time on the Hyper-V host, but this is a better answer, as the NTP usage means the network should remain on the correct time.