Medium should use a formula for payout that minimizes the incentive for “coordinated inauthentic engagement”. Formulas of this type exist.

The main principle is this:

The actions of one account only affect how the money from that particular account is allocated. The actions from one account do not affect how the money from other accounts is allocated.

In general: If a group of accounts all interact intensively with one another then the money that they gain is at most equal to the money that the overall group paid to Medium (minus some overhead).

If a person creates a new fake robot account then the additional money earned by the scammer is at most a bit less than the subscription paid for the robot account.

If you want details about how to design a good payout formula that reduces cheating then I will be happy to provide guidance.