I agree with your concern, as I understand it, but I do not think that what you are proposing will address that concern.

I agree with the concern that if one person or a group of people pretend to run multiple witnesses rather than just one, then that is a problem.

However, I do not think that your proposed approach fixes the problem.

• The minimum number of witnesses required on the various Graphene blockchains is only 11. So, I do not see how voting for 45 witnesses helps.

• Even if anyone (whether a conscientious stakeholder, or a deceitful conspiracy of stakeholders) votes for 45 witnesses (of which 3 are controlled by the conspiracy), then the conspiracy is still getting voted in. Although, perhaps to your point, it is true that their fraction of the total witness payout will be less than if a conscientious voter had unknowingly voted for 11 witnesses of which the conspiracy of 3 were coincidentally among them. However, a voter still does not know if all 11 or all 45 witnesses are colluding.

Maybe the introduction of some random number of witnesses into the pool would be beneficial. But I wonder to what end? If all witnesses are colluding, what damage can they do? If they do not report off-chain data correctly, or do not report at all on a particular event, then that will be obvious to everyone. Perhaps the biggest issues as I see them are:

(a) an on-chain game involving random number generation by witnesses, and

(b) an on-chain game where players must make a move within a certain amount of time but where witnesses could suppress a move by a player and thereby cause the player to lose a turn/move.

For this to succeed requires stakeholders who have less voting power than the conspiracy and/or stakeholders who do not have the means to detect these issues taking place.

I do not currently have the answers to these concerns, but I am glad that you are raising them.