What's missing in Joel's example is what makes multi-tasking worth while in computers: one
thread is blocked so the CPU is free to run another thread.
If you suggest managers only give humans one task at a time, they'll ask what that person is going to do while blocked? For example, we rely on services from partner systems. We often wait for them to become available or implement some defect correction.
Seems that a potentially difficult part of making one task work is to never be blocked. For this partner system example, one might mock out the partner until they come around.
Never being blocked is addressed in TOC with buffers, no? Does Lean have other nifty ways to handle it?