I’d be really interested to know how it works too!
In all likelihood, it’s probably some degree of glancing and estimating from the staff. In some cases, I’m not even sure they do that; in one instance logged in my spreadsheet, Vampire at Chessington was advertised at 5 minutes, yet it was nearly out the entrance and I waited 60…
I had an idea in my head of a near foolproof way of reliably estimating queue times. Imagine some sort of system that logs the amount of people entering and leaving the queue that’s paired with the ride computer system that logs the ride throughput, uniting queue entry/exit management and ride throughput management in perfect unison to make a wonderful, near foolproof queue time estimation system! Goodness knows how you’d implement it technically, but in the world of IoT and the like, it must surely be possible, at the very least…
It’s unlikely to ever come to fruition, but I can dream!