My 5-reference is 24 years older than your 3-reference.
A three-term recursion and the computation of Mathieu functions.
R. Lacroix coauthored with Felix M. Arscott MR0541916
Felix M. Arscott coauthored with František Neuman MR0283282
František Neuman coauthored with Árpád Elbert MR1167505
Árpád Elbert coauthored with András Sárközy MR0154023
András Sárközy coauthored with Paul Erdős1 MR0195837
And here's the weird story behind it.
Felix M. Arscott [wikipedia] was a professor of Applied Maths (and former RAF pilot during WWII). His lectures were never dull, not the least reason being that he had a funny way (to us students) of moving his eyebrows up and down as he spoke. One day in a first-year math course he taught for engineering students in fall 1977 he talked about his research on solution methods for ellipsoidal wave equations (which could represent sound waves coming out of jet engines), methods that were suitable for implementation on programmable calculators. My buddy and I, who were taking that course, thought it would be fun to try coding the method on our recently purchased calculators, but he had a TI 59 (with magnetic card storage) and I only had a lowly TI 58 that lost the program when I turned it off. So I went all-out with FORTRAN the following summer, coding it as my first serious program while I was still learning FORTRAN as my first "real" programming language. And after a while (getting past the usual off-by-one loop counter mistakes that novices make) Dr. Arscott started to get really excited about the results my program was producing, and he wanted to explore some parts of the solution space, which I obliged. The parts he wanted to see were harder to solve, causing the program to run out the 2-second CPU time limit for batch jobs fed into the IBM System 370/168 mainframe via the students' punch card queue, so he set me up with a timesharing account on his research budget. That turned out to be a very valuable thing to have when I switched to Computer Science one year later.
I don't know the specs for that IBM mainframe, but it was replaced by the more impressive Amdahl 470V6 which had a whole 12 MB of main memory and ran at the lightning speed of 12 MHz.
Anyways, after that summer in 1978 the electrical engineering office called me and my buddy, and handed us envelopes from Dr. Arscott containing prints of a math paper with our names on it, which allowed me to have a "publications" section on my resume that looks mighty impressive - and I never had trouble finding work since then, sometimes when I wasn't even looking.
But I was having much more fun in second year EE writing software than doing actual school assignments; doing things like creating a thermodynamics problem description language and solver, rather than just solving the thermodynamics assignment with pen and paper. Maybe I could have dropped out of school at that point and made piles of money in industry because such software didn't exist, but I wasn't thinking about making a living, I wanted to have fun with computers and the university was the only place I knew that had one. So it was absolutely a no-brainer for me to switch programs from EE to Comp. Sci for the next school year. I found myself surrounded mostly by students that entered the program seeking only to have a career that was in high demand, while I was there to have a blast.
During my first term in Comp. Sci. in fall 1979 my professor of Numerical Algebra (Hugh C. Williams - check him out on T5K) made sure all his students attended a talk by D. H. Lehmer (of LLR fame) at a conference hosted at our university. According to the MathSciNet reference the same conference also featured my paper, but I was oblivious to that until I found my Erdos number recently. At the time I thought Lehmer's talk was interesting but I didn't realize how important his work is to the field of prime numbers.
I remember Lehmer giving a photo slide show about how he used a machine made with holed discs and bicycle chains. When a light beam had a clear path through all the discs a photo sensor circuit beeped. On hearing the beep, he would shut off the motor which drove the whole contraption and manually turn back the wheels to find the number that had passed through. I don't remember exactly what he was doing with the machine because I didn't understand it then, but I think now he was looking for numbers that were simultaneously congruent to a bunch of residues modulo prime numbers. That would make sense, and I don't remember this detail, but he could probably detatch a disc and reattach it with a different starting rotation to use a different residue.