In practice you often get weird timeout issues and unintelligible errors in kernel logs, hard resets, drives disconnecting or worst, data corruption. It's also always difficult to tell if it is a failing drive, or if it's a controller bug, or power supply issue.
My 1U Raspberry Pi server with two mirrored spinning rust disks on USB-SATA has been up continuously since September and has been busy with running SearXNG, DNS, email, web (including dynamic sites), MySQL, and so on, while also being used to compile while heavily going in to swap.
In other words, it has been doing non-stop disk I/O since September without an issue.
This is because everything in the machine has excellent power :)
Some are OK, some are great, and some are garbage before they make it to a warehouse shelf.