I've observed plenty of flights that were, say, R>5 and stayed that way until one of the intermediate windows (e.g., T-72 or T-48) and then processed a batch of apparent CPUs. There may be "stalls", but I also suspect that either the sweep wasn't set to run for that flightor, perhaps more likely, Rcpu (the hidden bucket that CPUs come out of) was set to 0 during previous sweeps.