1.4 performance regression over time vs 1.3.2?

1.4 performance regression over time vs 1.3.2?

We're currently running a script-intensive app on jQuery 1.3.2. I did some quick testing last week against 1.4 and saw what in some respects is a perf regression.

The torture test case is switching a tabbed sidebar to very busy, dynamically-generated content. One big .html() write, a dozen or so small .html() inserts, and a whole bunch of event binding. No network access, animation effects or live events; no third-party plugins (including jQuery UI); testing on IE7.

On the first iteration, 1.4's perf is around 15% better than 1.3.2. However...

Our IE perf has always degraded over time relative to a freshly-started browser, but on 1.3.2 it was very gradual, with a lot of up-and-down noise along the way. On 1.4 it degrades immediately and almost linearly, to the extent that after only a couple of dozen iterations it's worse than 1.3.2.

I appreciate there isn't enough information here for a diagnosis. I was wondering whether there are any known performance gotchas in 1.4 that we should be avoiding, or whether 1.4 made any deliberate perf tradeoffs that might explain what I'm seeing.