I've noticed this as well. And it's not simply restarting the miner; I run mine with -r 240 so it should be restarting every 4 hours or GPU fail (I've not confirmed if this is happening in the logs yet FWIW). I got up this morning and hash rate had dropped and didn't come back until I rebooted. Bounced my machine and blamo... right back up to the 5130-5140 range.

Sapphire Nitro+ RX 580's - 6

Current Cal Hashrate (Nanopool): 6,300.0 H/s (or 1050 per card)
Claymore output: 5140 H/s (avg 856 per card)

Coreclock: 1380 (if I try for 1390 it hangs)
Memclock: 2250
Power: ~115w per card
-h: 1800 (default)
Temp: 58-60c
Drivers: Blockchain beta

Changing the -h flag in any way (other than the default of -h 1800) provides worse results or restarts. Attempting to set voltage control hangs/bootloops. I had to set my core clock lower with v10.x.

For comparison, when running 9.7 I would get would get average of 5009 H/s, ~129w /card, temp 62c with
Coreclock: 1425
Memclock: 2250
-h: 2030
Mem and core volt: 875
Power: ~129w/card
Temp: 62c

It really depends on how hard you want to push the card / what the extra power consumption is worth to you. I'm running:

-a 4
-h 1920
-cvddc 850
-mvddc 850
-cclock 1150
-mclock 2250

Average Hashrate for last 6 hours: 2,963.3 H/s (or ~493.9 H/s per card)

Runs about 80-85w/card and maxes out around 779w total from the wall (total rig - 6 RX580's, 1000wPSU + 850wPSU). Temp 45 (fan 60%),

I'm kinda curious what the flags are to get 700+ H/s though?!