Just about every system out there that has a synchronised clock service (Linux, Windows, Mac, Solaris, AIX, etc) all use NTP by name, or a variant of NTP called something else. On Windows, it's called w32time, and unless you're joining AD (if I recall correctly), it should be able to sync to a proper NTP server.
NTP is designed to work over the public internet and can synchronise to millisecond accuracy easily across a switched network. If it doesn't, there's something broken in your configuration, or you're attached to a server reporting bogus time and you don't have others in your pool to compare with (which is a problem with your configuration). 100ms should be expected across public internet, 10ms or better if you have a nearby clock source.
NTP uses round-trip delay, not one way time delay, as per RFC 958. It does not assume it's fixed for the transactions involved. Depending on your configuration, it may take some time for NTP to synchronise your computer, but in general it's fairly quick. A clock that keeps time to sub-1s over 2 days as Compton's did would synchronise so quickly as to be considered immediate for our purposes.
If by some chance you're not using something other than NTP as a time protocol, I suggest you change - it's about the only thing that regularly works across all modern operating systems.
As an ex-admin there, I can inform you that Deakin sync off Melbourne Uni's Stratum 0 servers. The publicly available NTP server was at Stratum 2 and routes out to the public internet via a tortured route that's often impacted by sudden changes in latency, congestion and the like. As you can see 20ms is easily possible despite that. Last I'd heard, they'd disabled public access to the time server though, so I might have to report you
Any bank that works to that standard now would be out of business, particularly since NTP can be implemented for peanuts and give millisecond accuracy. Banks are now implementing PTP, an even more extreme version of NTP, giving accuracies in the nanoseconds.
They use it for things like High Frequency Trading, and in that situation, they control every single parameter. For each customer allowed to access their HFT network, they'll put in a rack in their datacenter with the same length network cabling as every other customer to ensure there's an even playing field on latency. Then they synchronise using PTP to ensure they're all on the same time. A ex-colleague of mine spent time measuring RTT latency down to nanoseconds on cables for one of the major US banks, trimming copper off as appropriate until it met spec.
Pretty sure FT8 would work under those circumstances