Why does Proof of Stake need a source of randomness, and what is the catch about it?

Proof of Stake must randomly pick which validator proposes the next block, but a blockchain is deterministic and cannot produce true randomness on its own — so it needs a special decentralized randomness mechanism.

In Proof of Stake, validators are chosen at random to propose blocks (rather than competing by brute force). That raises a puzzle: a blockchain is a deterministic system — given the same inputs, every node must compute the same result, otherwise they could not agree — and a deterministic system cannot, by definition, generate a genuinely random number by itself. So the network needs a decentralized way to agree on an unpredictable value that no single participant can manipulate to get themselves picked. This is the same fundamental limitation that gives rise to the oracle problem: anything a blockchain cannot compute deterministically — randomness, or real-world data — has to be supplied by a carefully designed external mechanism rather than conjured internally.

Go deeper:

• Chainlink VRF — Chainlink docs — a verifiable on-chain randomness source no single party can manipulate.
• Proof of stake — Wikipedia — how validators are selected to propose blocks.