Anent the Smotherman paper. Woefully incomplete and inaccurate, but perhaps a starting point.
https://people.computing.clemson.edu/~mark/interrupts.html For example, he cites the CDC CYBER 200 as 1981. Nope--CYBER was mostly a relabeling effort. For example, the 6600 became CYBER 74, the 7600 became CYBER 76. The CYBER 200 was the relabeled STAR 100, which had its origins ca. 1969. The "Invisible Package" was part of the original design, It was large, contained all 256 registers as well as a potload of machine state data--and, as a consequence was very slow. It was very complete, so a job could be interrupted and restarted from the "drop file" which contained an image of the job's memory as well as the invisible package. Very useful for debugging, as one could have several drop files, each marking a point in a program's execution. I've never satisfied myself as to the actual reason behind it. The STAR, being very large, was prone to hardware failures and jobs were usually very long (hours). The whole scheme made it possible to resume a job after a system error had occurred. --Chuck