One aspect of maintenance is understanding the mindset of the original designer.
This is pleasant for very well designed machinery. Here's a long description of a Teletype restoration I did a decade ago.[1] Mostly this is about the mechanical processes, but sometimes I wrote about the thinking behind the design.
Teletypes were originally rented, with maintenance included. So they were intended to be reliable and repairable, and capable of a long life with periodic maintenance.
All parts outside the motor are individually replaceable.
Parts were treated against corrosion by Parkerizing, a chemical treatment involving hot caustic baths that leaves a rust-resistant coating. Few parts are unreasonably tiny, so you don't need tweezers and magnifiers. This did result in a bulkier machine than really necessary. Most moving parts are powered in one direction and spring return in the other direction. If something sticks, that doesn't cause further damage. Almost every screw has a lock washer. One of the few exceptions was due to a drafting error, as I mention in my writeup.
Mechanism design balances size, cost, ease of repair, wear, and lubrication requirements.
The number of people really good at that is not large. All the good Teletype machines were
designed by two men, Howard Krum and, later, Ed Klienschmidt. There were some other, inferior designs best forgotten. (The Teletype Model 26 was what happened when management wanted a
cheaper machine than the classic Model 15. Many Model 15 machines are still running; few
Model 26 machines are. And the Model 26 turned out to be no cheaper to make.)
Once you can appreciate this, you'll see good and bad mechanical designs more clearly.
It's clearer in the mechanical realm than the software realm, because failure is more obvious.
A lot of things have left me a bit flustered, to the point I assume the main reason it has been done like this is that parts where taken from a parts bin from the manufacturer and cobbled together to work.
There are a few examples on my car like that that are a bit infuriating (and were corrected on later revisions of the model, which I know because a friend has a later revision).
Mechanical design is just orders of magnitude harder than software or electronics. Those are 2D, and largely 1D, and in software there is no concept of tolerance, wear, or yield.
Teletypes were originally rented, with maintenance included. So they were intended to be reliable and repairable, and capable of a long life with periodic maintenance. All parts outside the motor are individually replaceable. Parts were treated against corrosion by Parkerizing, a chemical treatment involving hot caustic baths that leaves a rust-resistant coating. Few parts are unreasonably tiny, so you don't need tweezers and magnifiers. This did result in a bulkier machine than really necessary. Most moving parts are powered in one direction and spring return in the other direction. If something sticks, that doesn't cause further damage. Almost every screw has a lock washer. One of the few exceptions was due to a drafting error, as I mention in my writeup.
Mechanism design balances size, cost, ease of repair, wear, and lubrication requirements. The number of people really good at that is not large. All the good Teletype machines were designed by two men, Howard Krum and, later, Ed Klienschmidt. There were some other, inferior designs best forgotten. (The Teletype Model 26 was what happened when management wanted a cheaper machine than the classic Model 15. Many Model 15 machines are still running; few Model 26 machines are. And the Model 26 turned out to be no cheaper to make.)
Once you can appreciate this, you'll see good and bad mechanical designs more clearly. It's clearer in the mechanical realm than the software realm, because failure is more obvious.
[1] http://brassgoggles.co.uk/forum/index.php?topic=43672