Getting a good RNG on 1970s hardware is almost impossible without some kind of custom device. There is no end to examples of horrible RNGs from that era, sometimes manifesting in amusing ways. There was an implementation of Minesweeper that naively used the rand() call from C's standard library and generated utterly predictable boards.

I am rather impressed at how the author of the article actually tracked down the appropriate RNG function to verify exactly what happened. This is only possible because the PARC community was so small there would only be a couple of ways something like this could have possibly happened?