Wow, I'd like to live somewhere there's distilled water coming out the tap.

I can see parallel not working well. 

I have 2 C1 in series.  Two pumps in the loop.  Output of the first C1, hot, goes to a car AC condenser with various fans mounted I used while awaiting the arrival of a radiator for the second C1.  The condenser goes to a pump into C2 cold (pump sees cooler water), then output of the second C1 went to a stock radiator with the 3 stock fans and then to a stock pump and on to the cool intake of the first C1.  The second radiator arrived and since the condenser was cooling better than the stock radiator, I put the two radiators in series before the pump to the first C1.

So, I have solid cooling although I note two of the fans on the second radiator, an aftermarket, seemed to be running slow the other day.

The way you tried it was to load up two C1's with a single radiator!?.  Even if the load were magically distributed equally I don't think you'd get proper cooling in warm or  hot weather.

A less expensive pump, same manufacturer, gave me trouble with a crack along the interface of the top and bottom which I fixed with epoxy.  Then it still leaked but along an invisible crack on a side seam of one of the port tubes.  That one was difficult to seal.  Hot/warm water, slow leak, evaporated without evidence for quite a while but I knew it was leaking because I had to refill the tank.  It was that seam on the port tube to which one attaches a hose that was leaking.  Perhaps a strain on the hose while moving the miner triggered it.  Once fixed it was okay and has been running for months no leak.  I have a new pump just sitting here but since the leak was fixed I haven't wanted to shut down to swap out the pumps.  I use distilled water. 

I only refer to others posts - there were a string of complaints about the heat with no internet so I added the fan.  Haven't experienced it.

Yes, but aren't S5's known to sometimes overheat when the internet goes down and there's no hashing going on?  I personally added a rear fan and wired it to a full 12v so it's always running full out even if the internet goes down and the controller board signals slow or no fan speed.

Clean and reapply thermal paste?  I had a hashing board on an S3+ that started running hotter than other S3+ hashing boards and although I lowered the voltage and frequency and after a cold restart seemed okay was dead the next morning.  Still dead.  Signal missing on if I recall correctly, pin 13 of the ribbon cable.  If I had repasted immediately perhaps I could have saved the board.  On the other hand it might be that if a chip is going it's going to go repasting or not.

I'd guess that the approach with the greater aluminum mass and surface area is more effective but more expensive.

I just depress the reset while powering on, keeping it depressed 90 seconds, release the button and power down, wait 45 seconds then power up and I find the default stuff.

Have your tried a 90 second depression of the reset button?

Two C1's, cooling in series like this:
C1#1 -> radiator -> pump -> C1#2 -> radiator -> pump -> C1#1.
Pumps and radiator fans on a single separate adjustable power supply which failed.

I realize an improvement would be two supplies, each with half the fan/pump load with the C1's still in series.  That would provide a failsafe on either pump, or either fan/pump supply.

It would mean some fans from each radiator.  As it is I have a car AC condenser after C1#1 before the pump to C1#2 and had a single 3 fan radiator after C1#2 and before the pump to C1#1.  I got in a second 3 fan radiator and since the car condenser was cooling more effectively than the single 3 fan radiator, I put the two 3 fan radiators back to back in series before the pump to C1#1.  I'm thinking of adding fans from my shut down Saturn to the condenser.  So if I go to two supplies driving my C1s' cooling this summer, I could have one of  the two 3 fan radiators and the additional fans going on the condenser and one pump to one supply and the rest to the other.  Point being if the fan load on each weren't split and one supply dies then one C1 would be getting hot distilled water from the other albeit after passing thru a radiator or condenser.  

It was above 94° in my backyard shade yesterday so I dialed up the C1s' pump/fans power supply to 12.45vdc, perhaps higher.  Some time in the middle of the night the supply failed.  This morning one C1 was running one not, the one running had a highest temp of 80°C.  The two C1's are in series cooling.  Shut them down, swapped out the supply, and they are both up and running normal speed and temps.  They've only been  up 9 minutes so it might be too early to tell if they're both okay but looking good.  

My mistake, both were down according to the pool.  Perhaps a board hit 80°, hashing stopped but the recorded temp remained?

Odd a single x.  Having one supply on each hashing board when one supply completely failed I got a straight line of x's and it was chain 1.

I run my C1's at 243M/0675 for a hashrate about 983GH/s but one had dropped badly.  I thought reseating the connectors fixed it.  I had first seen some improvement by bring the voltage up to 0700.  I had digital readouts on the supply outputs then move the DVM to a pair of wires back from the PCIE so as to have the voltage readings the miner sees.  The supply seemed okay according to the DVM output.  The supply has two leads to the miner, only one of the two PCIE having DVM leads.  So, the hashrate had come back up but then dropped and stayed down.  I looked and one supply was dead.  That's been replaced and the miner is back running at 983GH/s.

I thought it was a PCIE connection problem and shutting down, adjusting the PCIE, then restarting fixed it for a few days.  Today a power supply on that machine died.  I replaced it and am hashing okay.  Time will tell.  It was a moderately warm day in the mid-80's and the power supply fan had failed.

Wonder if they're getting any returns having the stripped screw hold drilled thru and 4-40, washers and nuts in that hole.

So I've been using 360 watt supplies, one on each blade for S3's, s3+'s, C1's and S5, no matter works on all.  The supplies are modified with an added cooling sink to the bridge rectifier, a line filter, a 25 turn pot instead of a 1 turn pot and a digital readout.  This have typically a 14 gauge conductor pair to each PCIE-6.  I adjust to 12vdc.  I've  been changing the voltmeter to a pair of leads run to a PCIE-6 connector instead of right off the supply.  On a C1 the difference is .2 vdc on each of the 4 supplies so those had been running at 11.80vdc, same with the other miners more or less.  The difference being that when having both the first voltmeter and the second measuring at the PCIE, there was an approximate .2vdc difference between the two readouts.

I put a voltmeter on the Poweredge 750 watt.  It's stock voltage is 12.15vdc.  I calibrate all my voltmeters to a personal standard.  On a problem S3 which was running at 12vdc measured at a PCIE on each blade, I replaced the two 360watt supplies with the Poweredge.  I was sure I'd see less a drop in voltage under full load, the 360 watt dropping .2 vdc each.  Surprise!  The 750 watt Poweredge dropped from 12.15 to 11.8 under full load measured at the PCIE's.  That's a drop of .35vdc.  Even if I were to go in and adjust the Poweredge internally to 12.35vdc no load, that still shows a greater drop than the pair of 360watt supplies.  So, I have to assess the pair of 360watt supplies are stiffer than the single Poweredge.

Wait, I just get I'm comparing apples to oranges.  The .2vdc is the voltage drop due to the 14 gauge wire length while the .35vdc is the drop from no load to full load.  I'll have to fire up the two supplies I replaced to get a no load/full load difference.  But not tonight.

--------------

Falling hashrate problem with a C1.  It has been running at intentionally lower frequency to baby it.  Should be running at 960GH/s@243.5M and after a cold start it does but drops to 850GH/s after a few hours.  Up last night I restarted due to 850GH/s hashrate. 

This has the digital voltage readout measured at the PCIE connector rather than at the supply.  When the miner first fires up, before hashing, before fans run, pump on another circuit, the voltage at the PCIE measures 12.15v-12.17v and after hashing starts it's down to 12.00v so the inexpensive supplies drop ~.15v from no load to full load and there's the .2v drop on the lines so these supplies drop a full .35vdc same as the Poweredge. 

With the falling hashrate, when it drops, as each hashing board has a digital voltage readout at one of the two the PCIE, I see the voltage at 12.17v at one of the four hashing boards indicating it's not drawing much current.  Tells me exactly which board it being troublesome. 

I made a mistake building these leads.  Besides the PCIE plugs having a mix of heavier conductors on some and cheaper wire on others, after splicing the 14 gauge stranded to the PCIE wires, which I mistakenly cut overly close to the PCIE, I used heat shrink tubing on the splices and a larger dia. shrink tubing over all the splices.  This made the wires stiff near the PCIE.  I then put the miner on a cart but the wires came too close to the upper shelf and bending the leads flat put pressure on the PCIE board connections.  This is where I hope I'll find the problem.  If this isn't it then it's an ASIC that heats and accumulates errors (an internal board) but even when the hashboard isn't drawing much current while the other three are, there are no x's in any chain.

Did you use a PVC connector of any kind?  There is a special PVC used for hot water.  If one uses regular PVC, cold water regular stuff, it gets soft and perhaps dissolves.  The hot water PVC has slightly a brown hue.  Makes it interesting when connecting a new sink/faucet.  The size is smaller and one needs go thru transitions to get the hot water PVC to couple to the faucet.  Plumbers job security.

Cold water PVC also swells with hot water.  If you used the wrong PVC to route the coolant to a distant radiator, perhaps it got soft and developed a leak or swelled and reduced circulation.

And how will you restart?

Back when there was scheduled tasks one could put a hashmark before cgminer-monitor that ran every 3 minutes, save,  then send a halt command via ssh.  Without stopping with cgminer-monitor it would restart.  You might try /etc/init.d/minermonitor stop then ssh in and give a halt.  Then check if it starts heating.  Still you'd have a restart issue with linux shut down your ssh session would cease and not leave a reconnection. 

Living out in the hinterland I had been having internet loss problems some years back.  I put an X-10 device on the docsis cable modem and another on the first router.  On a linux system I have a program to ping my DNS every 5 minutes and test if the ping was good.  If not the x-10 would shut off power to the modem and router, time out a period then restart the modem, wait for its usual wakeup period then wakeup the router, wait for its usual wakeup period then test the ping again.  If left to the docsis modems own devices it sometimes would be down an hour mid-week early in the AM.  With this arrangement it's back up  much sooner.  But the internet losses have been much less frequent the last two years.

You may be able to run x-10 devices from a RPi or BBB.  I'm not sure the X-10 devices can handle miner current so you may have to run those to heavy duty relays.  Then when the halt command kills ssh connections and prevents reconnecting, you can ssh to the device controlling the X-10's, kill the power to the miners for as long as you need, then fire them up when ready.

No, I don't think that's the case.  The static option should be set on the miner.  The DHCP assigned addresses should be cleared from the router and allowed to be reassigned.  Typically 50 or 100 addresses are reserved by the router starting at 192.168.1.100 to be used/assigned to devices using DHCP.  I highly doubt the _router_ is assigning the .5 address.  The user would have had to go into the router and change the block of addresses from which the router assigns addresses.

That said, going DHCP worked.