GMail (AJAX version) is perhaps the ultimate test of ECMAscript-heavy performance. Here come numbers, with two connections: WiFi->cable->internet, ~2Mbit, ~60ms, and DUN->EDGE->internet, ~200kbit, ~450ms. (Pings are to the (nearest, one hopes) google.com.)

All tests run in portrait, windowed, with no other apps up, and CPU clocked at 400.

WiFi/MicroB:
00:51 cold to inbox fully displayed.
00:30 home page to inbox fully displayed.
EDGE/MicroB:
01:43 cold to inbox fully displayed.
01:28 home page to inbox fully displayed.

WiFi/webkit:
00:41 cold to inbox fully displayed.
00:35 home page to inbox fully displayed.
EDGE/webkit:
01:24 cold to inbox fully displayed.
01:21 home page to inbox fully displayed.

It surprises me that on a high-bandwidth, low-latency link, MicroB can actually win. (I reran that test; it's reproducible.) I can't imagine it's being slicker with the scripting, but that leaves the notion that it's more network-efficient when efficiency doesn't matter, but somehow losing that advantage when it should be more visible.

Major edit: Just figured out why webkit lost here, and why I like it better; it's using the full, themable, new version of Gmail! (Which has a light-on-dark color scheme, much easier on the (i.e., my, no flamage needed!) eyes, hence my preference.) So... not quite sure how to rope one of them into changing for a head-to-head, but I don't really care; webkit's still thrashing MicroB in 3 of 4 conditions, and that with a heavier load!