I knew a mainframe engineer and he certainly didn't have multiple spare mainframes laying around to play with. There might have been some form of partitioned install they called "test" but by the time people bureaucracy gets involved noone is "playing with it" and ergo, there's a lot less actual experts involved.
Bonus question: Was there never a firmware bug that would take down an entire mainframe? A certain server line was advertised with "mainframe like redundancy" on Windows, yet that's exactly my experience.