starting up a big lathe or mill spindle, or maybe a large phase converter with a heavy lever switch come close, but they don't scratch that "i'm in a super computer lab" itch as much as the fantastic whir of a raptor RAID.
utilitarian feature : you could hear a bad sector from anywhere in a large room.
The 15K rpm drives could sometimes be quite noisy; especially with 48 or more in an array.
Three of the engineers with whom I worked specialised in disk failure analysis; one of their tools was an inductive telephone earpiece pickup coil with suction cup, connected to a small amplified speaker. These guys would put the pickup on top of a drive and listen to all the electronics, spindle motor and head-moving voice coil electromagnetic noise as the drive spun up. This process would generally be followed by comments such as: "Hmm, head 4 sounds iffy", "This unit's on older firmware" and "Sounds like the heads aren't coming off the ramp"...
A lost/fading talent now that 2 out of 3 are retired". It was quite amazing to watch and listen to these guys at work.
but on the note of hearing differences in sounds, back in the day of dial up, since my ISP gave unlimited free dial up minutes, but it had to last no more than 2 hours (it would auto kick you off if you were on for more than 2) my modem would dial every 2 hours... And it got to a stage that I could know, by sound, if it was going to get a 56k link or a 33.6k link...
Ah, yes, the Mark I Real-Time Fourier Analyzer.
Though half an hour of clips is a bit excessive.