Follow-up Questions:
Thank you so much for your previous advice (see Part 1).

We had performed the copper/steel/stainless steel test back when the problem first surfaced several weeks ago, and it appeared (at the time) that the issue is due to moisture. We did not change the temperatures from our normal operating temperature of 1500°F (815°C). Should we adjust temperatures and rerun the trial?

This is why we took readings with only nitrogen in the furnace. Should we have done this? We had heard this from talking to one of the furnace companies. With only nitrogen in the furnace, the readings are very good (dew point at -50°F or below, and oxygen at around 25 ppm). These readings led us to believe that there was a leak somewhere in the mix line, but we cannot find it. We have probed every section of the plumbing with the flammable gas meter and cannot detect a leak. We have pressurized all of the line and cannot find a pressure drop. Today, we are in the process of pressurizing with helium and trying to detect a leak with a Varian tester. Previously, we had cooled down the furnace and replaced the muffle (we could not see a visible crack, but it was very warped). While the furnace was down for this change, we pressurized the cooling chamber and checked for any leaks there. We didn't find anything. I believe it was pressurized with 20 psi for a three- to four-hour period.

I forgot to mention one item last timie that we are finding very peculiar. It would seem that we can make the leak (or at least where it is being detected) move. If we check the furnace dew point and oxygen levels when the furnace is empty, we get very high oxygen about 1 foot after the transition from the heat chamber to the cooling chamber. If the furnace is fully loaded, the readings at the same area are good, and we do not see high oxygen levels until about 2 feet in from the rear exit.

If the operators load dunnage for about 30 minutes before loading parts, the dunnage will be very "rust" colored for the first 20 minutes or so and will gradually clear up. The parts will then come out looking good. Our typical dunnage load pattern is to load 6 feet of dunnage and then parts. The extra dunnage means that the dunnage actually starts to exit the rear of the furnace before parts can be loaded. If we stop loading parts or dunnage and move the dew point rod back in to 1 foot (300 mm) after the heat chamber, within 1-2 minutes of the last piece of dunnage or parts passing this point, the oxygen levels start to climb again! At this point, we have no idea what this could mean.

Answers:
Thank you so very much for supplying me with this new information. It is indeed an interesting problem. Here are some additional thoughts:

The fact that the furnace dew point and oxygen levels are low under nitrogen is a good sign. I read that you had run the furnace under nitrogen in your previous e-mail and did not ignore this fact in my recommendation but felt the copper-steel-stainless steel test would be more definitive in a dynamic condition. The copper-steel-stainless steel test should be done at around 1800-1850°F (980-1010°C) not 1500°F (815°C). The preventative maintenance data that you provided me shows what I would consider "normal" values. Furnaces do "breathe," so I am not surprised that the leak appears to be moving. If it is a leak, it is stationary, but the moving furnace atmosphere produces the effect you describe.

One of the most difficult leaks to detect is an intermittent water leak; something that opens up under certain conditions and closes under other conditions. If the piping is indeed free of leaks, this is likely what is happening.