Long before I worked, briefly, at JPL, MOPS, the Maneuver Operations Programming System was written for an IBM 360, in assembler. When they moved up to a Univac, it was re-written in whatever version of FORTRAN was current and worked very well. In fact, I'd bet money that that package is still running there because, like the old COBOL programs, It Just Worked. I know this because I had the privilege of working with the late
Dan Alderson, the last member of the team that migrated the package still at JPL when I was there.
In the almost three years that we worked together, I only saw him presented with a bug in the package once, and it turned out to be a user error. The user was trying to calculate the perturbations on Voyager I caused by 11 of the Jovian satellites, either to calculate their masses, or at least the maximum possible, causing the program to crash. It didn't take long for Dan to find the problem: the data for the satellites was kept in an array, and being written in FORTRAN, the array was designed for a max of ten objects. He suggested that the user edit a copy of the source to enlarge everything and use that instead of the regular program.