For the record, exchange below with @BenChung:
On a different matter: just came across your paper below on world-age in Julia,
https://benchung.github.io/papers/oopsla2020.pdf
The discussion in section 3.4 on "domain-specific generation" and on the associated "bottleneck" pattern is very interesting (*). I was aware of invokelatest but not of the mentioned pattern; the paper is five year old, is this kind of pattern still relevant? Any pointer towards updates (if any) on this topic would help [...]
(*) we've been systematically avoiding eval-uations in the parsing / code generation part of the OptimalControl.jl package not to run into world age issues
https://github.com/control-toolbox/OptimalControl.jl
For the record, exchange below with @BenChung: