Michael, I would be very interested in your test results. I still don't understand what is happening but the problem appears to have disappeared with drinking water that has the higher pH (7.8). I did check to see that pH was normal when 76 was mixed with DW and it came to about 8.0 so I don't believe pH of the DW is a factor. The drinking water batch was somewhere around 8.5 to 9.0, not enough in my estimation to make a difference.
The only other factor to consider, IMO, is the dissolved oxygen. But I have no way of determining if this is a factor or not. Surely the lower pH of DW is caused by dissolved CO2 so perhaps there is an abnormal amount of DO in the DW.
One other thought: is there any practical difference between mixing 76 with tap water and then storing it in an HDPE gas permeable container versus mixing 76 with water that has been stored in an HDPE container and subsequently stored in a brown glass bottle? It seems to me that the developer will oxidize in either case. And, perhaps, there may be different levels of DO in DW depending on how fresh it is, that is, how long it has been sitting on the grocer's shelf since it was processed/packaged.
I have easily seen the difference between leader material that has been out for days and the portion of the leader that is exposed when loading the cameras. I haven't measured it precisely but there appears to be about a stop or two difference, the long exposure time portion of the leader has indeed gone into reversal. My observations have only been based on that portion of the leader that has been exposed to daylight for the few seconds during the loading process.
When I get the time, I may mix a batch with the DW and run some comparative tests as well. I may end up losing another gallon of 76 and I am not the least bit sure I will uncover anything more than I already know. Plus, I don't own a densitometer so my crude spotmeter approach would be insufficient in this analysis, however, Bill's clever test worked well when using the spotmeter. Also, I was able to detect 3 stops difference in leader density between Rodinal and 76 when using my crude tool.
Ideally, I would like to solve the problem, implement a solution and be able to use Xtol. I do prefer it over 76 but I have never been able to keep it alive. I even tried Suzuki's solution with brown wine bottles and a Vacu-Vin device but it still died on me and I ultimately attributed that failure to the bad water.
EDIT: I have mixed another batch with DW and will let it sit several days and try an A/B test with the batch that was mixed yesterday with water stored in a PETE container. Maybe we'll see a difference there.
Something worth considering, which I believe someone else mentioned earlier on in the thread, is that since we never measure the amount of exposure the leader receives, we can't really conclude anything about the resulting developed density. What if it receives enough exposure so that it enters the reversal region of the curve? How do you know what exposure will lead to d-max?
I can tell you for example, when plotting curves using increasing, known exposures extending to the point where the shoulder is flat, I have sometimes had those densities measure higher than the leader density.
In any case, it is good Fred seems to have solved his problem. I would say however, that teacher you had in school taught some bad, meaningless lessons with his negative density "criteria".
For the fun of it, I've mixed two new packets of D-76 (same batch #) at the same temperature, one in tap water, one in some unopened distilled water from the local drugstore, and will try both of them on some Tri-X. What will this prove? Zilch since the distilled water isn't the Kroger stuff Fred used, but I figure why not try it anyway. I've done this experiment before with XTOL and some other developers but not D-76 or ID-11, so why not. I'll be pretty surprised if the sensitometric results are any different within experimental error. But you never know.