The linked article is working through some of the numbers on a pure math basis but the real world is messy in ways that article doesn't consider.

Most notably, if a drive has failed there's a reason for it and a lot of the possible reasons will be shared with other drives in the array. Was there a manufacturing issue and all of the drives are from the same lot (pretty common). Is the RAID in a hostile environment (heat, vibration, bad power, etc.)? Heck, is the RAID normally very lightly used and going to have heat problems if it's under full load for hours (days?) during a rebuild?

There are also factors like how the RAID controller is going to handle another read error - will it drop a second disk if the drive reports a failure? For that matter, if an array drops to "degraded" due to an error are you going to immediately replace the drive or are you going to write it off as a one-time fluke and let the array rebuild? Do you keep a pool of spare drives around that you'll drop the failed one into after testing? I've regularly seen an array drop a drive due to something transient, then rebuild onto it and not have any more problems for years.

Even with a 2-drive failure in a RAID5 you're unlikely to lose much data - almost everything is likely still there on the disks unless there've been catastrophic failures (e.g. an array of the old "Deathstar" drives which were prone to head crashes). You just may need to do recovery which will generally mean imaging each of the drives and doing recovery based on working with those images instead of the original drives.