We should straighten out out concept of fractional reserve.
In money theory, the fractional reserve is really the loan to deposit ratio, it defines a multiplicative ratio. For the sandbox, in the S&L technology, this is the mis-match between the typical sequence of deposits and the loans events. In the S&L technology, this is about 3% to 15%, and mostly a bound upper and lower limit. This is the float liquidity at risk, needed for price discovery.
The bit error stream is the S&L pit boss' share of that float risk.That number is likely to be .5%, very low. It is the ad cost when you look at facebook kitten videos; in the sandbox, the bit error volatility should be priced like ATM fees in the current system, except much cheaper. Most of the fractional reserve risk is borne by traders betting one set of inside information against some competing side. both innovative and thus bear most of the pricing risk.
So we get this mis-understanding of fractional reserve in current banking system, I mentioned it before. The basic S to L math still applies, but some of the reserve is hidden in regulatory mandates deposits held elsewhere, or three way trade regulations on capital. In other words, it is a very lousy implementation of a sandbox, but it is a sandbox.
These semantics go out the window for specialized trading pits configured for hedge funds and exchange funds, and these may be three color. Redneck systems has multi-color pits in the reference architecture, so some pits have these variable, and high cost ledger fees.