If you flashed the 0.4-2 release of Pali's u-boot and it half-bricked, the issue is that the bootloader somehow causes kernels to not recognize the presence of the 256MB NAND device at all. If you try to flash it from within the device, it'll probably stall after not being able to find that chip. This doesn't have anything to do with where you load the kernel from - if you load the power kernel from u-boot, it actually loads the kernel itself from the MMC partition.

There might be a way to turn off whatever is causing this from within the bootloader itself (didn't Pali say that it had something to do with OneNAND support?) but I haven't found this yet...

(Now, if only I could find a place which will repair an N900 USB port... I don't have the soldering abilities to do it myself, not to mention that I'd have to take the USB port off first, and finding somewhere/someone to do this in the US has gotten incredibly tricky...)

Edit: Wait, the problem is with the boot.scr? If that were the case, shouldn't "run attachboot" (which *should* ignore boot.scr issues) start up the attached kernel properly? (Although if you're using Power kernel+CSSU-Thumb, you'd have something that would probably crash shortly into the boot).