Even then, can you guarantee that the person behind the account isn't being paid to say things they wouldn't otherwise say?

Does it really matters if a comment is made by a troll/shill/bot or just a normal illinformed individual? The real issue is people believing inaccurate or false statements, which at the end may result in bad state policies. We can't solve this problem by checking identities.

Before considering technical implementability, we may want to consider if such a thing would be wise.

That guarantee isn't required for better online discourse, which is good, because it's not something we can reasonably guarantee.

Why should anyone ever make that assumption?