You should also read Kahan's commentary on the book: https://people.eecs.berkeley.edu/~wkahan/EndErErs.pdf.
Also, note that since the publication of the End of Error, Gustafson has given up on the original variable-length unums described in the book and switched to "posits" (or "type III unums").
He also goes over the history of floats and numerical computation, highlighting how often the best method was never adopted, for reasons as simple as patents or it not being better by enough.
Unums are better by enough, and they're being used at LLNL in experiments. It's just a hard ask for people to switch the number system they use, even switching from decimal to binary was a huge slog in the early computer age.