...
We assume the cost of running the matcher for the same data sent/produced to a channel is the same (negligible) regardless of what the pattern is on the corresponding for/consume. We also assume that the (not tight) upper bound cost of running the matcher to get a single successful match on a channel will be as if all the produces on that channel that eventually reduced according to the current replay log already existed in the replay rspace, and we ran the matcher over all of them.
...