I think we have pretty clearly established that there is an intermittent issue with the magic URL redirect functioning incorrectly. While perhaps not 100% established, my impression is that the database is probably always being correctly updated with the proper data regardless of browser. I agree with
IB-Dick's doubts that database scaling is the problem, and I think the experimenting with cutting the "assume read" window from 10 days to 5 days is not going to get us anywhere.
The important thing to analyze next is how the magic URL redirect is implemented. Not being a web technologies expert, I don't even know anything to guess other than a URL rewrite rule. Is that how it is being handled?
Whether the answer is yes or no, I think it is a safe bet that the redirect is handled by different code than that which handles the page generation. The page generation seems to work perfectly, so we should focus our attentions on the rewrite rule (or whatever code handles the redirect).
The error might be because:
- the magic URL redirect uses an incorrect database query (in some cases);
- the magic URL redirect pulls up old data (in some cases);
- the magic URL redirect produces an incorrect result, despite having accurate data (in some cases);
What else might it be?
"In some cases" seems to be a common denominator, so at some point we will need to ask "why is the error intermittent?" There could be some faulty browser identification code, but we should postpone that discussion for now.