Despite their ability to expand to large sizes using parallelism, later to be called clustering, they were in the category of "mini-computers".
Tandem's strategy was the emerging concept of "continuous availability" for applications requiring near-total uptime, in which single points of failure were minimized through mirrored storage devices, controllers and software, and operating systems engineered specifically for fault tolerance (using the Guardian operating system, later NonStop Kernel or NSK, and later NonStop OS).
[1] In 2022, HPE NonStop platforms underpin many of the largest banks, casinos, retailers, telephone companies, email systems, and stock exchanges worldwide.
Many high-uptime computer systems that were produced after the Tandem NonStop platform rely on a pair of concurrently running, parallel but independent processes.
Each pair of CPUs share hardware ownership of controllers and disk drives.
Writes to the mirrored disk pair are performed on both the primary and backup drives.
If one CPU in the pair fails, the other will take full ownership of the disk drives and tasks it was undertaking, providing continuous availability to that data and continuous operation of the system, including re-driving disk I/Os.