Great deep dive, I've been actively curious about some of the results you found that present themselves similarly in infra setups I run or have run previously.

Great deep dive, I've been actively curious about some of the results you found that present themselves similarly in infra setups I run or have run previously. This kind of miffs also: > AWS doesn’t describe how failure is distributed for gp3 volumes I wonder why? Because it affects their number of 9s? Rep?