On the Windows CE front, for the past week and a half or so I and some others have been working through a number of high priority issues for Firefox on Windows CE.  Conveniently, one of the biggest problems is startup time, so there has been some good synergy there between this work and the overall startup effort.

Given that web browsing is a primary use of many of these devices, and that due to their low power nature, they are almost always suspended rather than fully shut down, we can pay a portion of the startup cost during the machine's boot process, and then have fairly quick time to get a window open.  I have a prototype of this in bug 509249, and some next steps that need to be done there.  The rest of the startup effort will help as well, since the cost is often IO related.

In addition to that, one of the biggest issue is just general stability.  We're a fairly complex app, and working in an ARM environment is helping us shake out issues within our ARM Javascript JIT and any problems in any other areas, such as data misalignment (of which Windows CE on ARM is a lot less forgiving than any other platform) and issues in floating point libraries and fp emulation.