I don't think the right solution is to track pages read better. That would make the scam a bit harder (because scammers would have to either manually turn pages or fake the client software), but is still pretty easily defeatable.

A couple of things that Amazon could do to fix this behavior:

1. Instead of putting all KU subscriptions in a big pot and then dividing the pot among all authors by page read, divide the income from each individual subscriber among the pages he/she reads (so a read from a user who reads less would effectively be worth more). That way spammers who join collectives to read each others' giant books are just taking money from each other instead of stealing from the larger pot.

2. Don't pay out revenue for pages read until 2-3 months after the fact, and if a book is determined to be spam during that time withhold all revenues. This admittedly affects legitimate authors as well, but maybe Amazon could soften the blow by contributing the spam-attributed portion of the KU revenue back to the author pot so all the legitimate authors get a nice little bonus for their patience.

Note that either/both options might require renegotiating contracts with KU program authors, but if the spam problem is as bad as it sounds I can't imagine most legitimate authors objecting.