Another downside of the "throw another mutex at it" is if you're caring about performance(which you probably are) ti's a great way to introduce overhead via dea

Another downside of the "throw another mutex at it" is if you're caring about performance(which you probably are) ti's a great way to introduce overhead via death-by-a-thousand-cuts. Each mutex you take has a not-trivial cost but doesn't show up well on a profiler since the individual call is small and spread across an invocation path.