It's the early 1970s, and in this mainframe data center, the operations supervisor is a by-the-book kind of guy.
"He created a 'halt book' for the corporate programmers to document halts and actions to be taken," says a pilot fish who remembers.

In other words, whenever a job stops unexpectedly, the operator can look up the error number in the book and follow the prescribed procedure to fix it.

"The corporate programmers implemented a weekly report, and assured the operations supervisor that all expected halts were documented in the halt book before flying back to the corporate office," fish says.

"The next week, it came time to run the application, and the comptroller was waiting patiently by the printer in the fourth-floor computer room when the operator said there was a halt on the console.

"The operations supervisor insisted that the operator go ahead and follow the halt book, and explained how well corporate had documented all halts."

The operator checks the halt book -- and refuses to follow the directions.

So the operations supervisor and comptroller check the book themselves:

Reason for halt: You should never get this halt.

Action to be taken: Go jump out the nearest window.

"Needless to say," fish reports, "the programmer responsible for it flew back and was correcting the problem by 6 a.m. the next day."