USB thumb drives are prone to fail. Some models tend to still mount, sometimes read-only (ro) other times read-write (rw) with input/output errors when executing a simple ls command.

Exemples in journal for a faulty drive:
kernel: FAT-fs (sdc1): error, fat_get_cluster: invalid cluster
kernel: FAT-fs (sdc1): error, corrupted directory (invalid entries)
kernel: FAT-fs (sdc1): error, fat_get_cluster: invalid start
kernel: FAT-fs (sdc1): error, invalid access to FAT (entry 0x00000001)
FAT-fs (sdc1): error, fat_get_cluster: invalid cluster chain (i_pos 0)
VFS: Lookup of '\xe2\x95\x9a\xcf\x83\xc3\xbccf\xc3\xaf.6z2' in vfat sdc1 would have caused loop

It would be great if updatedb would have got a mechanism to detect these kind of errors during its execution and then skip the faulty disks/partitions instead of reaching an endless loop and filling up all available disk space:

# ls -lh var/lib/mlocate/mlocate.*
-rw-r----- 1 root locate 1.1M Oct 21 00:00 var/lib/mlocate/mlocate.db
-rw------- 1 root root 9.6G Dec 30 20:11 var/lib/mlocate/mlocate.db.PRvfsw

and never ending its process start :
# ps -ef | grep updatedb
root 3249 1 99 Oct22 ? 213573-14:47:11 /usr/bin/updatedb

The current available updatedb.conf options as well as updatedb (mlocate) version 0.26 leave room for improvement to let Linux (in this case Arch) systems stay up and running.