Use iperf3 on both ends to rule out optimizing results prioritization for speed tests. Watch network stats on both ends to also track retransmits, fast retransmits, lost retransmits, recoveries, locked sockets, etc... Run tests on different ports and test with and without using a VPN such as WireGuard. iperf3 will show losses and retransmits and will allow setting limits on number of threads. Different http protocols can mask or exacerbate these numbers whereas iperf3 will expose the raw traffic stats. Different congestion control algorithms and queue disciplines can also mask or exacerbate numbers. I would also be interested in seeing the stats for civilian vs. active military controlled accounts to see if they get priority which I would expect given low latency is required for drones.
You can just use Steam, they know what they are doing at Valve
Speedtest to ground segment (never really hits the internet) vs speedtest to closest (to the ground station) Cloudflare server.
Wouldn’t people who were downloading multi gig files be suspicious?
What's suspicious about watching a Twitch or YouTube stream?
I interpreted this as:
"Wouldn't people who were downloading multi gig files [become] suspicious [of the claimed speed because it would be obvious to them that it is not correct]?"