I've been running some measurements on bandwidth needs and the effects of latency on query performance against Postgres 8.3 (waiting for management to approve upgrade) on Windows. SELECTs of various quantities of rows and some INSERTs with bandwidth of 25 Mbps, 5 Mbps, and 1 Mbps, with latency of 50,40,30,20,10.
For selects of 1, 3, and 250 rows the performance was basically the same with each bandwidth, and the same effects of latency. However, for a select of 7100 rows, the 1 Mbps situation had identical performance across each latency value. Additionally, while 25 and 5 Mbps performance was linearly correlated to difference in rows (28x rows, 28x time), the 1 Mbps queries are at about 44x.
So, I was curious if Postgres has some built in throttling that would mask the effects of latency. This may be more of a networking question, but I don't think so. Seems to be specific behavior to larger queries on slower connections.