BTW: I am loving this debate! IT is exhillarating.
I believe the point is for the light NOT to fail, not service after the failure. Although it is great to hear that those two provide excellent customer service. I actually thought good customer service was dead; apparently not true at Halcyon and Salvo.
However from a strict engineering standpoint, if it can happen it will. If there is no penetration in the head or housing it will not leak. If there is an opening with 12 o-rings there is still a possibility that it will leak eventually. ( I know I am getting ridiculous, but I am pretty stubborn about making a point....ok just stubborn in general.)
Anyway the best solution IMHO is no switch at all.
In industry reed switches -with a solid state relay- and/or Hall effect sensors are used extensively in highly un-defeat-able safety interlocks where if someone opens this door you will be injured, and maybe killed.
The reed switch itself cannot take the kind of currents we use today. The old Darrell-Allen (big yellow clunker) used a large one. I have a vintage Darrell-Allen light, just for a museum piece, and have never used it as it would take some work to get it going, but it is cool looking. Mine is neatly beaten up, but in very good condition for a 30+ year old thing that is well used.
Anyway, I really enjoy this debate. I think the bottom line is:
listen to everybody's opinion then do what ever you want. I think if I were building another one for me, I would go ahead with the reed with solid state relay, but if I were building one for you I would ask what you want.
I'd say go for it, carry a back up, no matter what switch you choose. The fun for me is trying different things. If you made it once, you can make it again. If it was ever made, it can be improved upon until it is perfect.
Please post pictures when you get something done.
Here are pictures of mine.
http://picasaweb.google.com/wes.newperspectivesinc/LEDDivelights#