The speed of light is a design constraint

Light in optical fibre covers roughly 200km per millisecond, about two thirds of its speed in a vacuum, and that sets a floor under the round trip between any two places that no engineering removes. Ian Gorton puts New York to London at about 28ms one way through fibre on the direct path, so over 50ms there and back before routers add their share, and real paths are longer.

Anything that waits on another region pays that floor on every operation: a synchronous write, a consensus commit, a quorum spanning regions. The only way to lower it is to put the copies closer together, which makes them less independent.

That is the tension in choosing locations. AWS separates the Availability Zones within a region by a meaningful distance, so that one storm or earthquake is unlikely to hit two, yet links them with fast private fibre so that traffic between them stays cheap. Across regions the usual compromise is asynchronous replication. DynamoDB is one example: its global tables propagate updates between regions asynchronously, and Gorton notes a write can take several seconds to appear elsewhere.