I've found lots of examples of outstanding physical performance under a vegan diet, but I've been unable to find examples of bold theoretical breakthroughs being made under a vegan diet (the closest example I could find was Ramanujan, but there were other things about Ramanujan that suggest that there's absolutely no way most of us could live the same life and then end up in the same place), and my own experience has been really discouraging. After about 10 days on a vegan diet, regardless of my energy levels or legible performance metrics, I'll pretty reliably stop being able to, or stop being interested in progressing original ideas.
There are lots of possible exits here: Vegans and inventors were rare until very recent history, we shouldn't expect to have many records of people who were both, we might end up with none, even if there's no relationship between those things. It's also fairly likely that the effect would just be a result of a creatine deficiency, or a choline deficiency, or something like that, which can be fixed with a supplement (although Dr Gregger doesn't recommend it due to the incidence rate of high contamination in creatine supplements, but maybe you can find a brand you trust.)
(Anecdote on choline: I have a vegetarian friend who, at some point, stopped thinking in the sort of focused/precise/fluid/driven way that our project needed. They speculated that it might have been because they were on choline initially, then stopped. So I said yeah, I noticed a change, try getting back onto the choline? I don't think they ever got around to it. The lack of concern that they showed is actually one of the things that bugs me about this, because I also experience that, when I'm low: It kinda seems like we have be on the supplement to hold onto an understanding of why we need to stay on the supplement, and as soon as we lapse we forget what it was like, and how important it was? (Depression also seems to work this way. A depressed person often cannot imagine or remember not being depressed.))
There does seem to be a consistently replicated (edit: creatine findings not consistently replicated, actually!) finding that supplementing creatine enhances memory and intelligence in vegans. (and doesn't for omnivores) (baseline cognitive performance in vegans is similar to omnivores, but personally I'm kind of expecting it to turn out that to be a result of vegan-leaning demographics starting on a higher base, then being lowered)
But even if it can be fixed with a supplement, I don't think most of us are taking creatine and choline! We probably need to have a conversation about that.
Previous discussion of health and veganism, when they touched on concerns about cognitive impacts, generally failed to allay them.
But you can also just judge each breakthrough separately, conditional on what they had access to. If they're deep into a problem past where anyone has been and then go further, that might be more impressive, but it may not be, in case it's easy to identify the next (possibly hard) subproblem after solving the last subproblem. So you can approach it locally/greedily, without thinking ahead much to where you need to go, only about where you are now and the next step. I think upper-year and grad-level pure math and theoretical computer science problems can be like this, although maybe not as hard as you're asking for.
Something harder I have in mind would be something like having a non-local/non-greedy approach to solving a problem, where you have a major breakthrough just to get a sketch of a proof or to come up with possible lemmas, and then it takes further breakthroughs to close things up. If your sketch is wrong, then all the work can become basically useless, and you don't progress things for the next people to try (except by ruling out a dead end).
It's like thinking more moves ahead in chess with multiple hard moves to identify, compared to just making the same number of hard moves to identify in the same game, but never part of the same sequence simulation in your head. Both are multi-stage, but the second one is local/greedy and isn't more impressive than making the same number of hard to identify moves across games, fixing the total number of games played.