Thanks for mentioning that. Especially because that reminds me of something else (14 years ago) that nobody taught me:
On a remote videoserver I had to make sure that the application I wrote would make an administrative video backup to a removable harddisk and I could be certain that neither the UUID , not the serialnumber, nor the label was predictable with a drive swap.
I then digged through the /dev/disk directory and I ended up using the location of the drivebay: /dev/disk/by-path
Funny backstory:
b.t.w. the server would have had a more than satisfying uptime if it wasn’t for the time the whole datacenter went down (which took out many racks and damaged many servers despite being hooked on two different electricity nets , both backed up by generators). 
And the time that the company moved it to another datacenter 100km eastwards.
Not that it mattered anyway: If you have about 700 days uptime it might be a good idea to bring it down anyway for securitypatches (which, ofcourse, is a good opportunity to update the OS since it is down anyway).