[dTal]

Well, yes and no. The amount of RAM consumed by the filesystem driver is negligible compared to the truckloads of filesystem data shoveled through it. If we assume that errors are comparatively rare, the code itself is unlikely to be affected. Even if you're unlucky enough to get RAM corruption in the 0.01% occupied by the ZFS driver, the chance that a bit will flip in just such a way as to make a checksum succeed when it should have failed due to a second bit flip is virtually nonexistent. Much more likely that it simply crashes in some way. As such ZFS is much more resilient to on-disk filesystem corruption from bad RAM than systems which don't do any checksumming at all.