> I still feel like it may be possible,
I don't think so. The timer hardware has fixed and limited functionality, more complex issues are to be handled in software. I usually try to pull out tricks using DMA, but it's "forbidden" here, and the combined slave-mode modes like they are in the newer timers, are not available in TIM3.
I can imagine using one timer as the gated clock source, routing its output externally to ETR input of other timer in external clock mode 2, where the slave-mode controller is used for reset (i.e. input signal goes both to first and second timers' CHx pin, to be input to the slave-mode controller - this signal maybe can be routed internally). We spent 4 pins and it still "suffers" from what you've identified above:
> it'll trigger as soon as the pulse reaches the threshold, as opposed to when the pulse terminates.
and I don't quite see how that one can be fulfilled.
JW