The plan for unwinding central bank QE, at least as stated by the Bank of England and the Fed, is that asset sales won’t start until the recovery has got to the point where sales won’t have to be reversed.
That means waiting until the recovery warrants interest rates being raised sufficiently far above the floor that any need to loosen again can be achieved by lowering rates, rather than restarting QE.
The reason for this was that central banks did not want QE to be the marginal tool of monetary policy. Partly because of worries that stop-start-reverse sales would disrupt bond markets, and partly because of the extra uncertainty in using QE relative to the more tried and tested interest rate tool. [Though one might contest this relative statement since there is much debate about the effectiveness of very low interest rates near their floor, and we have accumulated experience with using QE].
The curious thing about this exit plan, in retrospect, is that it suggests an apparent inconsistency in the way that asset purchases were used on the way in. If central banks had really wanted to avoid QE being the marginal tool of stimulus, they ought to have bought so many assets that it was possible to keep interest rates sufficiently far above the floor that the interest rate tool could subsequently be used to fine tune, as news about the economy evolved. [You might ask: isn’t this just what the BoE did, keeping rates at 0.5%? No, is the answer: that floor was chosen as the point at which the monetary policy committee then judged that further cuts would be contractionary rather than stimulative, because of their effects on bank balance sheets, a point revised down subsequently.]
Past policy can be justified by noting that there were also serious concerns about the credibility of central bank policy with bloated balance sheets, so they had to weigh balance sheet size against avoiding using QE as the marginal tool [though I don’t recall this argument being made back then]. But in future, now we know that QE does not make the world fall apart or lead to hyperinflation, if the ‘marginal tool avoidance’ argument survives at all, it can be applied symmetrically.