Play-to-Earn with real world earnings will never be sustainable given the state of the art in reinforcement learning.

The RL game playing agent essentially becomes a form of NFT mining that no human could ever complete against.