>> ... The way MTBF figures work, if you have two devices both operating
>> at the same time you're roughly TWICE as likely to have a failure ...
>
> Which is not exactly the point.  The point is the likelihood of a [complete]
> cooling failure.  That would go up roughly infinitely if you had two units.
> Further, there would be no need to "change at TBO" - you could run until
> failure.  As for check valves, I can envision a couple of ways to build a
> "dirt simple, dirt cheap" valve that would be more than sufficient for our
> purposes.  Ed or whoever just bought a toy mill could start a cottage industry.
> With check valves, one could switch water pumps periodically in flight, thus
> keeping them both "exercised" and promptly detecting any failure.  As for high
> demand moments, if the pumps were plumbed parallel, you could use them both for
> takeoff, climb, etc.

Yes, for all the reasons you listed, especially the last, my preference is two pumps in parallel. Unfortunately, _I_ do not have a mill, toy or otherwise, and do not plan to buy one, so if anybody has a way to manage this I'm most interested in hearing it.

> ... Viola' a dirt simple, dirt cheap check valve.

I can think of a few problems with this arrangement. For instance, at full flow, the two pieces would be forced together, essentially forming a single "vane" cutting the flows in half. No problem, but if you reverse the flow, what forces the vanes apart allowing the flow to force them completely flat/open and thus blocking the flow? OK, slight engineering change - if you hammer the very tips into a small angled lip this would allow reverse flow to "catch" on them and force the halves apart. They wouldn't seal as well, of course, but better than nothing.

You also have to deal with the heat of the coolant - again, careful epoxy selection would be a must.

The commercial versions seem to use either a flap or ball, with a spring to assist the return. I'd almost rather spend my $200 per valve for something so critical. God forbid those suckers detach, flow down to the radiator, and completely block the flow from BOTH pipes.

If we're talking about elegance here, I'd almost rather see something in "Y" form where the valve is a butterfly between two inlets and combines the flows into a single outlet. Then only one valve is needed. If only one pump is pushing, the butterfly valve would block reverse flow into the other pump. If both pumps are pushing the butterfly valve would open relatively evenly (assuming equal flow from both pumps) and allow both to flow.

Regards,
Chad