This might be related to Limewire -- I don't know. Usual setup: 4.something Limewire on XP SP2, using Sun's JRE 1.5.something.

Every so often the entire machine spazzes out: everything becomes slow, and network-using apps become unusable (take forever to make a connection, fetch a Web page, whatever).

Task Manager shows one of the svchost.exes using 95%+ CPU. This is with a fairly beefy Athlon -- and it stays that way for several WHOLE MINUTES, on a 2GHz CPU. That's around a trillion clock cycles being used by, apparently, NETWORK SERVICES. The only NETWORK SERVICES I can think of that need a trillion clock cycles are really heavy crypto or hundred-K-connection Web servers, which I'm not running. (Limewire is limited to 30 simultaneous downloads and 5 uploads -- yup, broadband.)

A Google for svchost CPU usage problems found numerous complaints where the problem was some service getting stuck starting up; the services control panel in XP showed nothing of the sort during several of these incidents. There were only two network services started, and these were "started", not "starting", and several that were stopped. The running ones were DNS and some RPC thing.

Event viewer does show a bunch of application events generated whenever the seizeup starts, and a bunch more when it ends, but they aren't listed as coming from Limewire. They are "Database stopped" or "Database stopped the instance" after the seizeup, and "started" or "started the instance" before. They come from various alphanumerically named executables like "wuclault" and such, whose names are unpronounceable. Googling them shows them to be Windows Update related. Turning off automatic updates hasn't stopped the seizeups, however, and they started when I got Limewire 4 and started using it regularly. Plus, as others in this thread have noted, Limewire is quite the resource hog.

The system is clean of viruses and spyware. The seizeups don't have any particularly regular timing, but have an elevated probability of occurring if I'm actively doing stuff in limewire, particularly just after a "find more sources" button click. Naturally, this results in it finding no sources, since clicky->seizeup->Limewire starved for CPU and all network stuff getting painfully slow during seizeups->Limewire hangs with a nonresponsive UI and drops all its connections and can't make new ones, so no sources are found.

(And I can't even go and do something else while I wait for Limewire to recover. Web browsers will spin for 30 or more seconds before even beginning to render a page, news can't be fetched, mail can't be fetched, and so forth, while the seizeup is in progress. Everything just tends to spin or timeout, and games and such won't run acceptably with svchost hogging the CPU either! So much for "multitasking" -- another great idea that falls flat when anyone actually tries to implement it. :P)