That's where I came across them in around 03/04, though the company was moving away from them at the time. Two intel processors somehow kept in lockstep, allowing either one (or its RAM) to fail and be replaced with no interruption to the system.
the originals ran like 3 'cpu groups' or whatever they called it...
all instructions ran on all 3 and the results were compared - one of the cool things done with this was taking faulty 'cpu groups' out of service without down time
This 'voting' strategy is used a lot in hi-rel aerospace and military computers as well.
And in space. Eg shuttle used 3 main dSpace controllers, instead of hardening the hardware. Still cheaper, and much faster.
'Aerospace' usually covers 'space'. The shuttle used 5 IBM AP-101 flight computers (4 synch'ed voters for the core avionics software, 1 for independently written backup software). The AP-101 was designed to be a hi-rel computer for military use, so it's very much 'hardened' (including EMP protections), but not internally redundant. They were not cheap. Ken Shiriff has several articles about the hardware: https://www.righto.com/2026/03/ibm-4-pi-computer-history.htm...