Don't Reuse Bonds in Proof of Stake
Most proof of stake systems use the security bonds, the "stake", for 2 unrelated goals.
(1) The first goal is to stop consensus forks. If validators should double-sign, the stake gets destroyed to punish them. In order to be able to punish, the bonds need to be locked up and innaccessible for a long time. Locking up funds is expensive by the interest rate. These security deposits only need to be 3x bigger than the amount of money transfered since finality. That way double-sign attacks destroy more money than they can steal. Ethereum and Tendermint each lock up about 1/20th of the coins, so these interest rate costs are very expensive. In Flying Fox the bond can be relativly small, like 1/5000th of the market cap, so the interest cost on this bond is about 250 times less than ethereum or tendermint.
(2) The second goal is to stop soft-fork censorship. Ownership of stake determines how much power you have in deciding the next block of the chain. This group has the power to selectively censor transactions to modify consensus. This group need to be sufficiently invested in the system so that they will make decisions that are in the best interest of the system. Since it is spendable, it isn't so expensive to convince people to participate. This bond needs to be very big. Bigger than 9/10ths of the market cap hopefully. Luckily this bond never gets slashed, so it can be accessible all the time. For example: if 2/3rds of the bond holders worked together to censor any transactions spending from a target address, then it is the same as if the money in the target address was deleted.
If we use the same bond to accomplish both (1) and (2), then it will do both things poorly. It will be too big for (1), and it will be too small for (2).
Flying Fox mixes the delegation mechanism with the channel mechanism, which is a way to create the second type of bonds.
If you split proof of stake bonds into 2 types, then it can be secure with much lower fees.