We were "IO-bound" using a setup that executed a maximum of two requests at a time, completely blocked processing new requests until earlier ones fully completed, and took a long time to execute requests because of unexpected latency in a dependency, while 20 frustrated users sat in a room hammering the refresh button and dumping more tasks onto the queue.

You're absolutely right when you say there was something seriously wrong with the setup! The problem is that it was more or less what you'd get if you set up a default Rails app behind Passenger, and Passenger did this because some of the major Ruby gems weren't threadsafe (e.g. Rails, until 2008), which comes back to concurrency problems with the ecosystem.