When you set the PPS for a game, you're setting the rate at which ship positions update. The game uses those packets to measure loss. Those packets are given serial numbers, and the game knows how many to expect at what times, since they're sent at a regular rate. It looks back over the last 100 packets (which is about 3 seconds at 30 PPS) and counts how many are missing -- that's your downstream loss number.
Ping packets are send out once per second. They have timestamps attached, and the replies come with those same timestamps. So the ping displayed means you really did just send a packet and get a reply, and it really did just take that long. Those numbers are pretty solid.
Ping packets also include the measured downstream loss (from the other guy's perspective), which is where your upstream loss number comes from.
One of the glitches with the system right now is that the numbers only update when you get a packet. If the traffic goes totally dead, the upstream loss meter is never updated by incoming pings, and the downstream loss just cheerfully waits for the next (long delayed?) position update to arrive, and doesn't see any holes in the sequence in the mean time. So the system doesn't do a good job of alerting you when the connection has gone
really sideways.
I've seen it behave oddly at other times, too. Sometimes it'll spaz out and count up to 100% loss, even though the other guy is clearly and smoothly flying around. I think it's getting confused about where it is in the sequence? Regardless, it seems to sort it out pretty quickly.
I haven't seen skipping with no loss reported (other than what I mentioned above), but I don't doubt that others might have. If there's some sort of intermittent interruption or delay in the connection, that would do it.
No doubt the system could be improved.
That teleport was pretty nasty. With an intermittent lag spike up around 1000 ms . . . something that bad can happen.