Terrestrial latency is bounded by physics: light travels fiber at ≈200,000 km/s, i.e. ~1 ms RTT per 100 km of actual route (which follows roads and railways, not the straight line). Add a few tens of microseconds per device crossed.
Useful benchmarks: Paris–Lyon ≈ 4–6 ms · Paris–Marseille ≈ 10–12 ms · Paris–Frankfurt ≈ 8–10 ms · Paris–London ≈ 7–9 ms · Paris–New York ≈ 65–75 ms · intra-metro ≈ <1–2 ms.
Design consequences: synchronous replication requires under ~2 ms (hence <100 km); chatty applications (many round trips) amplify every millisecond; and moving the service closer to users remains the only optimization physics allows — hence choosing your data center by latency to your users and clouds.