I now seat the spring closed keys as carefully as the stack keys to eclipse the light 360 degrees (or to the same pull of the feeler gauge) with just the lightest pressure regardless of how strong the spring is set. My reasoning for doing so is shown in the picture below. The illustration of course is exaggerated to convey the concept. In reality it happens on a much smaller scale. A .001" change in a pad's alignment can be enough to create a leak.
Hi John
As I said, I agree that increasing the spring tension to make a pad seal won't last as long as a good sealing pad with normal tension. I guess it's a matter of degree that would affect how temporary it is. How uneven was it before increasing spring tension, how much was the spring tension increased, how much does and how deep does the tone hole press against the pad, etc. How fast (or slow) the pad will start leaking depends on all of these things.
I also "seat" the closed pads in the same way. I imagine they would seal against light with even less tension on the springs, but the springs have to be set to a certain tension to seal, just like the fingers.
My reasoning for doing the same as you that pads harden over time. They start somewhat soft to reduce noise and to overcome their own slight uneveness. The more accurate the pad is initially, the better it will keep this accuracy when it slowly hardens and loses its ability to that incredibly small accomodation it needs to do.
I would say pads are instlaled to an accuracy of about 0.01mm or better, because if I put my 0.015mm feeler between a pad and tone hole then the leak light shows what I consider a "leak" that is more than I expect to overcome by a finger or a spring.
The pads that I have seen where spring tension was increased to overcome inaccuracy were mostly on new instruments where it was not possible to do anything about it for budget reasons. It is sometimes possible to reduce spring tension in these cases, but not as mcuh as would be optimal. Since the pads sealed with this spring tension and budget was very limited, it was preferred to leave it (and fix whatever was more critical on the instrument). In these cases, I have very rarely seen the problem shown in your drawing. Maybe it happens more when the the difference between pad inaccuracy and spring tension is bigger, significantly more than most cases I've seen (just guessing).