I beg to differ. It works pretty good most of the time. I agree to have backups for sure for anything you don't want to lose just in case. But yeah, do try to fsck first. It's saved me a lots of work several times.

(Of course with gentoo, you'll do anything to avoid a reinstall!)

In the beginning of hdd fixes, Microsoft DOS scandisk tried to scan the magnetic surface for errors and rewrite in different area for corrections. The problem is that error correction of a single byte is easily fixed by checksum, but many bytes of error can not be fixed by FAT table info. So fixing errors from two copies of FAT table(front and rear of a partition) is a disaster.

Windows scandisk technology is now totally improved. It is done by finding files that are corrupted, reinstalled that file from the compressed library of the operating system. Once the OS is whole, the FAT table(front and rear) is rewritten. The data you have may be corrected, but maybe also lost. Your computer however can be operated without fuss.

Fsck is like dos scandisk, used to fix file systems. As file system improve, fsck may lag behind. Fsck has to be used with synaptic file manager to reinstall some files from repositories, or from a dvd/cd source, needed sooner than later.

The problems are generally generated from FAT table on hdd; not having more space for more than six fragmentations of any file. Drupal used to have this problem where apache and mysql had difference of checksum, fsck may save the day. Gentoo compiled kernel files(except config files if changed freqently) generally do not have to be reinstalled in the file system, if it is not fragmented in the FAT table.

When fsck tried to do it on the whole file system; it may be much better to just reinstall file by file pending on checksum fidelity alone.