Since October 6, 2008, the Fed has been paying interest on reserves at 0.25%. Apparently this institution is now permanent, but the Fed has no experience in "normal times" with setting the interest rate on reserves along with its fed funds target rate and the discount rate. Of course, in more advanced (only kidding, of course) countries, Canada for example, central banks have been paying interest on reserves for some time. Since 1994, the Bank of Canada has imposed no reserve requirements on chartered banks, and has been paying interest on any deposit balances held overnight with the Bank of Canada. From an economist's point of view this seems like a good thing - efficiency has to improve, as the inflation tax ceases to distort holdings of bank reserves, and there is no rationale for reserve requirements, other than as a means for generating revenue from the inflation tax.
Bank of Canada operating procedure, in "normal" times, is to target an overnight interest rate (the counterpart of the fed funds rate), with lending at the "standing liquidity facility" (the counterpart of the discount window) at the overnight rate plus 0.25%, and interest paid at the "standing deposit facility" (reserve balances) at the overnight rate minus 0.25%. In normal times, overnight reserve balances are essentially zero. However, since April, 2009, the Bank of Canada has set the interest rate on standing deposits equal to the overnight rate, at 0.25%, and has targeted the quantity of "settlement balances" (i.e. reserves) at $3 billion. Given the size of the Canadian economy, we might take this to be the equivalent of about $30 billion in bank reserves in the US - a substantial quantity, but about 1/40 of the current level of US bank reserves.
Could the Fed just adopt the "channel" system for setting interest rates that the Bank of Canada uses? Maybe. Maybe not. A key difference between the US and Canada is that central bank intervention in Canada happens through the LVTS (large-value transfer system), the key financial payments system, which has only 15 participants. Lending in the LVTS system is essentially riskless. In the US, the fed funds market has thousands of participants, and lending is unsecured and risky. The reported fed funds rate is actually a weighted average of trades in only a small segment of the market. Thus, in the US, it is not clear what the relationship should be between the fed funds rate and the interest rate paid on reserves, in "abnormal" or "normal" times.
In the medium term, the Fed seems to be concerned with "locking up" reserves, as the fed fund rate rises. The idea seems to be that if the banks start to attempt to shed reserves, this would drive up the price level (lower demand for reserves, and more outside money held as currency). Since these reserves are roughly equivalent to short-term Treasury bills currently (they yield roughly the same rate of return), one solution (provided we accept the Fed's premise) could be to pay interest on reserves at the T-bill rate. Banks have to be willing to hold reserves at the T-bill rate, and this provides a direct link between what would otherwise be an administered interest rate and a market-determined rate. This approach seems superior to proposals to have another class of reserves - essentially term deposits held with the Fed. Since the term deposits would have to bear an interest rate higher than the T-bill rate, on average, pegging the interest rate on reserves to the T-bill rate would give the Fed a higher return on its portfolio - critical in the current environment.
In the longer term, the US could join Australia, Canada, Mexico, New Zealand, Sweden, and the UK, by eliminating reserve requirements. Someone may correct me on this, but it appears to me that the Fed has the power to do this. The Fed can change the required reserve ratio at will, and could set this to zero. Then, when times are "normal," the Fed could adopt a channel system like Canada's. I'm not sure where to find this data in a readily accessible form, but an eyeballing of the time series of the reported fed funds rate and the overnight rate in Canada shows that the variance in the first series is remarkably larger than for the second. Probably the Fed would need a larger target band (bracketed by the discount rate on the high side and the interest rate on reserves on the low side) - perhaps 75 or 100 basis points, rather than 50 as in Canada.