@Douglas Young yeah your propagation math is right — 0.13ms is essentially irrelevant at that range, you're not fighting physics, you're fighting the connection interval scheduler.
The dominant latency source on the STM32WB07 stack is almost certainly the connection interval. If you're running with a default CI somewhere in the 45-100ms range, your round-trip floor is already 90-200ms before you factor in any retransmissions. At 30-40km you're going to have a higher packet error rate than close range, which means the link layer retry mechanism kicks in more often — each retry burns another full connection interval, so your average latency will be higher and your variance will be wider compared to shorter links.
The other thing worth checking is whether you're seeing any throughput-related queuing delay. At 2Mbps the raw pipe is fast, but if your telemetry bursts are hitting the notification queue faster than the CI can drain them, you'll see latency creep up as packets stack. Worth logging queue depth if you can.
If you share what connection interval you're currently negotiating and roughly what your packet error rate looks like at that range, I can probably help you narrow it down more specifically — there's also a firmware discussion started a while back in the Dev Hub around CI tuning for the WB07 that had some useful numbers in it.