I coded some further optimizations to ligaments, so that reloads of my mainline rookery (which is now >2200 pages) are slightly snappier (as short as 2-3 seconds for some changes). But I still need to sit down and really think through the incremental reload story, as recompiling large rookeries is likely to be my most common use of Rheo going forward. This could be a place where Rheo really outshines native Typst, for example, as the latter is not optimized for recompilation of fragments littered across a repository which are synthesized into a single Typst compilation context as they are being incrementally developed.