There are plenty of articles out there about the chances of compromising a seed phrase via keyword duplication. The fact of the matter is, as you have noted that the number of options is very limited - this is due to the BEP39 open source wordlist function (or similar) and despite what you may think, word / phrase duplication and the ability to bruteforce is very real. I've actually brute forced a BEP39 seed phrase by accident ffs. If by convention you think of how huge the likes of Rockyou and the dirb lists are and the continued success rate of those, it's less probability you need to be worried about in terms of seed phrase security and more about direct targeting vectors. Think about the probability of a hash collision in anything below SHA256.....really low but still happens and can still be manufactured. I have encountered some APTs now able to manufacture hash collisions in some algorithms which I thought was absolutely impossible. Seems not. I think there is a lot to be said for the improvement of how crytpo assets are secured and I am always thinking and considering new ways to improve cryptographic security functions. I think there is a lot to be said for the way that Solana have introduced state proofs into their cross chain protocols (on the Sol side) which bears equivalence to quantum encryption in some aspects of the mechanism which is pretty cool. Anyway, I'd be really up for working on tweaking and playing with security mechanisms for the TON ecosystem if anyone else is interested.