Former protocol developer for Tezos here. Once you get to 33% of the stake you can stop the chain, so the idea that you could control the RNG at 50% isn’t that

Former protocol developer for Tezos here. Once you get to 33% of the stake you can stop the chain, so the idea that you could control the RNG at 50% isn’t that scary by comparison. You can only influence the RNG if you go last. That is to say, you could not completely predict or control the RNG unless you control 100% of the stake. At 50% of the stake, you go last 50% of the time, and even when you go last you don't have the ability to fully control the RNG (although you could do a pretty good job at 50%, at substantial expense to yourself). The only means anyone has of influencing the RNG is by not revealing their random contribution, which they're punished for by losing their staking rewards. (The original tezos whitepaper had a higher penalty, I believe you forfeit an additional "safety bond", but I don't know if that made it into the implementation.) The hope is that incentives like these are enough to discourage that behavior. All cryptocurrencies depend on incentives for security, in part for cases like these. But this is a particularly bad case where someone with a sub-33% stake could attack the chain. Fortunately, at <33% control of the stake, the damage you could cause from that attack is small and the attack would be very expensive, so it’s not considered a major problem. I think the worst you could do is censor some transactions for a time, and deprive some others of their staking rewards. (Although I was a protocol developer, I never worked on anything related to this, so don't take my word as gospel.)