In this post.
Additionally, some of Judy's space and speed improvements come from strategies that could be used by any data structure but aren't because they're not good software engineering.
Seemed like an odd statement. That's like saying "Assembly is obtuse and it's bad software engineering to use it." IMHO, it depends. If you are writing low-level code that will run very frequently, perhaps good engineering would suggest that you should optimize in assembler.
I think, there are places in your code where you need good runtime performance and spending programmer time and sacrificing generic dogma are justified.
Like the author says at the end. If some tool/library/service meets my needs (in this case robust and fast being among them), then I'm not too concerned about how it gets there:
Of course, you can certainly take the attitude that Judy performs well enough, and the fact that it's 20,000 lines of code doesn't really matter, as long as those lines of code work and you never have to look at them--and it appears Judy is mature enough that this is the case.