Backblaze regularly releases failure rate statistics of their drives, and it’s often a big enough dataset to be quite meaningful. I haven’t been keeping up with it lately, but there certainly was a period of time where there were substantial differences in the failure rates of different manufacturers.
So while you do still need to have drive failure mitigation strategies, buying more reliable devices can definitely save you time and headache in the future by having to deal with failures less frequently.
That’s wild. Did you need a special program to parse stuff out of the data stream? I guess it would mostly come in as http reaponses, so it wouldn’t be too hard, but still an interesting problem.