Latency also depends on the methods used to route the packets around among the satellites, to decide which satellite (or satellites, plural) to send the packet(s) to, whether there are any or will be any close to the intended destination, how many satellite hops it needs to make[0], and more.

In short, it's complicated, much more complicated than I can see any of the articles discussing, which is why I raise the question. All the articles ever talk about is speeds, which is worrying.

[0] You need LEO to get low surface-to-satellite latency, but then each satellite can only see a comparatively small part of the Earth's surface, so you need more hops to get the packet(s) to the right satellite. I remember working on a system where the engineers helpfully(?) doubled our radio range, effectively halving our system-wide throughput. It's experiences like that that make me worry about these things.