[...]the only reason people might use died is by (incorrect?) association [...]
Not the only reason.
My gut instinct is that, strictly speaking, "future present" die is "correct" above, and the only reason people might use died is by (incorrect?) association with the hypothetical (but now ruled out) past scenario (at some point before time of utterance, I did make that assumption, but I no longer do).
In the title, you mention a "now discarded past." You may inform us you have discarded the past hypothesis, but it is still in play semantically because you have presented it for our consideration. The initial point of reference here is the moment when we all may learn the truth. One take on that would be some sliver of time after your death. So died is reasonable. That only shifts when you say "think". Thinking demands a shift to some time before death. But there's no need to recalibrate the original point of reference.
Died also gets the point across that dying has been completed. Interestingly, a search of "would go when I died" got about 30,000 hits, and "would go when I have died" got none, as did "would go when I had died".
I found a related discussion on Word Reference breaking-the-backshift-rule-for-indirect-speech ...
Keith Bradford's answer there seems particularly on point.
Borrowing one of his sentences, I think it supports my idea about the time point in the point of view that I expressed above. Two of his samples were these -
- We knew yesterday that John will call us when he arrives tomorrow.
- We knew yesterday that John would call us when he arrived tomorrow ? (... this seems awkward because it has knowledge conflicting with doubt.
I would add the following variations -
- We knew yesterday that John would call us tomorrow when he arrives.
- We knew yesterday that John would call us tomorrow when he arrived.
Five and six are acceptable to me, but not four. Five sounds like John is going to call from the airport. Six sounds like he's going to call from the hotel once he's settled in.