The diamond norm categorises the single-use distinguishability of two quantum channels. The operational setting is the following: you have access to an unknown unitary, which is promissed to be either $U_1$ or $U_2$ and the goal is to distinguish whether you are in the case where you have $U_1$ or in the case where you have $U_2$.
Is there an efficiently computable measure for the setting of distinguishing channels when we are allowed to make two or more uses of the channel? In other words, is there an algorithm for efficiently computing the maximum probability of success for identifying which channel we have, given $k$ uses of the channel, for any $k$?
I encountered a related question on the tensor product of many copies of the unitary. However this assumes that the copies of the unitary are applied to a probe state in parallel. I think it does not capture all strategies that may query the probe state in sequence, or use an adaptive strategy based on measurements. Moreover the discussion there is only about an upper bound.
It seems that this is a more difficult problem. Harrow et al (2010) found an example of a channel that can be distinguished perfectly with two uses in an adaptive strategy, but for which no finite number of uses without an adaptive strategy would be sufficient for perfect distinguishing.