Sorry, this is a bit crazy but...
Untaring meego kernel (480MB of source), MyDocs gets corrupted as ** (1h of 0 responsivity before reboot, prolly swappo), run fsck... waiting... last assuming EOF message appeared an hour ago, still keeps 100% processor usage, seems to be working. 26h later, sorry but ctrl-c. Trying from Ubuntu, after 2h after last assuming EOF message: cannot create unique file name. Grrr... few more tries later - sht will have to flash eMMC. Vista comp shows up, ok, lets give it a try, 14 minutes later touch testfile in MyDocs works. FFS really? Know FAT is win domain, but seriously... Only way to fix - WIN? (except format I guess) This is not right

HI
i had similar problems , i did a check from windows xp sp3 and now everything is ok. In the past mu n900 will take 30 min to read all my pictures .
Plug the n900 select n900. properties and disk scan . it work for me

Another lesson learned: getting rid of 50,000 files (out of ~200,000 untarred) in WIN takes 5 minutes, getting rid of the remaining 150k in shell about 30 seconds. All this doesn't make sense

EDIT: Funny thing is: properly untarred archive got around 37k files, how did I get to deleting 50k at around 40mb (out of 200mb) untarred on device??? Maybe the multitude of '8x questionmark in rhombus.3x same' that popped up while fscking is the key...

You need to use the IO Improvement Swappolube tweak and swap on SD.

hmmm... thanks, but from my experience the random restart at end of heavy task is exactly the result of having swappo on. Compiling heavy items with swappo results in random reboots, without finishes after hours but without restart and corruption. Can you give me some settings for swappolube IO tweak to test? Compiling DCSS should be a good test enough, gets to full swap/tmp very soon and restarts with swappo defaults in few hours, completes in few hours+3(4) with defaults, with no reboots. If there is better swappolube setting for compiling this should be mentioned/noticed/advertised even