Did you check if hda and hdc got swapped by accident? I once had a mainboard which had some awful bug: it swapped controllers after a warm-boot (i.e. reboot). Controllers were somehow presented in another sequence and Debian detected them the other way around. So: hda became hdc, hdc became hda.
However, mdadm does (AFAIK) not make assumptions about device names, but makes it's decisions based on the UUID and an event counter of an array. In any respect, you're situation is very confusing and the last thing a server administrator wants. Imagine a mailserver storing mail on an outdated disk after a reboot: it'll directly outdate the other disk too! -- To UNSUBSCRIBE, email to [EMAIL PROTECTED] with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]