The challenge then, as with essentially every language that doesn't use persistent data structures _as the default_, is having to recursively convert to and fro

The challenge then, as with essentially every language that doesn't use persistent data structures _as the default_, is having to recursively convert to and from those persistent data structures at the edges of your system where you need to interact with third party libraries that likely won't be consuming/producing those data structures natively. Unfortunately this is where the value proposition of a non-standard data structure library usually breaks down for general usage. That recursive translation layer sitting at the edge of your system is generally enough to completely blow up your performance budget and many times leave you worse off than if you had just used the default mutable data structures and made naive shallow copies everywhere in the first place. The best case study for this is JavaScript, which by now has a fairly substantial following of functional programmers, and has an excellent persistent data structures library in ImmutableJS, but the vast majority of functional codebases in JS just end up doing shallow-copy-on-write using the default mutable data structures regardless, because of the difficulty of taking proper advantage of the performance benefits of persistent data structures in an ecosystem that overwhelmingly deals in mutable data structures. That said, the fact that C++ is not a garbage collected language and the associated developer ergonomics costs of having to manually clean up your own shallow copies might skew the value proposition a bit, but given that if you're writing in C++ you're probably in it for the raw performance in the first place, it's unlikely to be enough to tip the scales.