I thought I was making something simple and immediately realised I was a fool. But at least I now know exactly how much of a fool I am.
Longer explanation under the cut for the gamedev curious.
I'm making something where the structure is that you choose paths that, for lack of a better term, 'recruit' various characters into a pool. You basically end up making three choices, at which point some pair of the characters will interrupt whatever would otherwise happen 'normally'. On reflection, I realised this means that unlike what I had thought at first, this creates a lot of branching. Because I need to be able to account for any path a player might have taken from this point.
I can't have an ending reference a character the player hasn't picked up. But I also don't want to have any ending that doesn't bring in previously recruited characters. This means I have functionally made a structure where every individual path needs a bespoke ending.
This is slightly less bad though than having a completely discrete path like in the top two diagrams. Because I can re-use the interrim steps and those are 100% the same. Plus those sections are then being seen much more. One of the biggest problems with the top two styles of structure is that each person sees only a fraction of the work but it takes no less time to do. With the bottom three, people are seeing a lot of the middle sections (eg. B-I in the bottom right diagram). Even a player who replays just once is seeing most of what's written there.
Also the endings of each do have re-use, because each of them will have a section that's the same only for a pair of characters to interrupt and derail into their ending.
Lastly, for anyone curious if I add a third choice between H and I in that last one, it'll mean 10 new passages.... which I might do. Not ruling it out, will see how much I get done by the deadline.