Hey Arthur
I've done a few tests around how latency affects insert performance and the results are about what you would expect, IE when you batch multiple statements into a single statement or transmission you are affected far less by latency. For an example in my tests @100ms (intercontinental latency) it took around 17mins to insert 10k rows via sequential insert commands (such as those currently used by EF) but only 7.5s to insert those same rows when using batched multi-row inserts (ie approx 135x faster). With a more reasonable 10ms latency it took a little over 2 mins for sequential inserts and 1.6 seconds for multi-row inserts (ie approx 75x faster).
Have a read of my article on this which goes into more details around the results.
