It was around 62-65C when I found miner. If it was getting into the 70C range I would have backed frequency back down.

Part of me wishes to believe that it is a thermal issue with temperatures running 42C and 43C at 200M, however when frequency is at 250M-300M temperatures are right at 50C with numerous HW issues.
That makes me think something else is wrong.

I can tolerate it hashing at 650GH/S with lower power consumption, but it is definitely not preferred.
I guess next question is does checking thermal paste void warranty (if not already voided) and any other thoughts as to what to check besides thermal paste.

I have an issue with one of my S5's that has ran the modified cgminer 4.9 firmware and been overclocked to 375M for around 2 months. (Other is doing very well but underclocked now)
It had numerous hw errors when I checked it (like 50%) and seemed to be not hashing at all a few days ago. It is in a utility closet in the garage, but temps stayed within 50-60c range 95% of time.

I have troubleshooted the problem and right now I am running at 200M stable (with a reboot as I write). As far as I can tell it is not just isolated to one board as both seem to having same issues, when I bump frequency up of more hardware errors,instability,no hashing etc.

So far I have changed power supplies, changed pools, ran stock bitmain firmware, reseated all cables, and looked for noticeable signs of heat damage.

I am thinking of breaking down board and checking thermal paste next, but wanted to get some responses as to the next logical steps... (Will probably have no luck with warranty)