As usual the answer is 'it depends'. I have r9-280x Linux rigs I can leave 24x7 but some rigs with the same os/gpu cards and just a slightly different model of motherboard I find I need to re-power (as apposed to reboot) them every 18 hours or so. So there are no hard and fast rules, just general guidelines for high availability, of which here are some:
a) Choose a motherboard with reputation for mining stability.
b) Run a linux based miner (Bamt/cgminer works for me).
c) Use a high quality power supply and don't run it close to it's limit.
d) Make sure you have failover pool configured in your miner.
e) If overclocking introduces instability dial it back down.
f) If undervolting introduces instability dial it back up.
g) Use cards of same model per rig where possible.
h)
Keep everything as cool as possible (Trade this off with overclocking/undervolting based on your situation).
i) If your doing it at home, use different rings of your mains for each rig.
j) Constant power cycling, overclocking or undercooling will all reduce the working life of your rigs dramatically.
Hope that helps...
best answer... almost i can contribute though and make it best here
c)use this.These test use proper EQ not some multimeter toys.
http://www.realhardtechx.com/index_archivos/Page541.htm also searchgoogle power supply review list for more.
g) i woulndt be so sure. Why u so sure?
k) if you can, move your rigs to basement/coldest broom in house and cover all windows os there is pitch blackness for max coolness.