Arduino nano randomly resets- and requires new program to recover. Weird, right?

Has anyone had this occur? I haven't seen this myself and I haven't assessed the units its claimed to be happening on, but I am hoping to get there and check it out myself soon. I want to be armed with any relevant knowledge beforehand though if possible, so here I am.

I built 20-some prototypes which basically just turn on an SSR to regulate a heater. Most are working pretty well (especially considering my shoddy programming skills). Two of them recently have started to "erase their program" after some time running. This is the client's assessment not mine- I would assume that behavior isn't possible, but she says that when she plugs into the module, it lights up like normal, but there is no status on the serial monitor (and there should be- its part of the program). But she can "fix" it by re-uploading the program to the board, and then operation resumes as normal.

Anyone heard of this or similar behavior? Thanks for any help!

Do they resume functioning if they're powercycled? If they're manually reset?

Capture a failed unit and read out the contents of the flash via ICSP - see if there's any difference between what's seen there and what's on a working unit.

Is the power supply stable? Could there be a glitch on the power rails?

If you can, post the schematic for us to examine.

Sorry for the late reply, I don't get email notifications so didn't know there was a response! They are power cycled with each use, a master switch (analog spring timer) powers up assorted associated devices. (5v power supply, blower motor, attached SSR, and external hour meter). The power supplies have been very stable always reading exactly 5v. Here's a semi-primitive schematic: http://i.imgur.com/PzXAPgT.jpg

T represents the thermistor temp sensor and is located remotely along with an LED.
I still haven't figured out what happened with that unit in particular but the others all seem to be working pretty well.