To answer that question, you'd have to look at the legal history of Ponzi schemes. The Madoff ponzi is a good example.
The trustee responsible for returning funds, Irving Picard, had to determine how much money was left to be distributed, and how much to return to investors. Neither has a simple answer.
Determining how much money is available is tricky because the gov't has the power to compel those who benefited from the ponzi (the early investors) to return their profits for redistribution to the investors. Picard focused on the larger recipients, such as the owners of the New York Mets, who were big investors with Madoff.
Since bitcoin payments can't be reversed and the investors are mostly anonymous, that's not possible for the pirate ponzi. But it is possible to subtract withdrawals from the original investments of each investor, which mostly accomplishes the same thing.
In the Madoff case, the courts decided that returns should be limited only to actual funds invested with Madoff, excluding fictional paper profits that never actually existed. Since the profits Madoff reported to his investors never actually existed the courts saw no reason to try to refund them.
My proposal for payb.tc is to pay off investors using the same methods the courts determined for the Madoff case. In short: take the original investment, subtract any BTC that was actually paid out, and then pay pirate's refund in proportion to their net investment losses.
If pirate comes through and delivers the full amount due, then payb.tc can distribute according to the actual account balances, which would be great.