https://gitlab.com/sylva-projects/sylva-core/-/merge_requests/7492 introduced $all_dependencies_cache to memorize dependency chain across units. A similar pattern can be applied to unit-def calls, which are currently re-computed on every call with no caching. Each unit-def call: - Merges unit_definition_defaults + all unit_templates entries + the unit's own definition via merge-append - Runs interpret-inner-gotpl over the merged result This is expensive and redundant for repeated calls on the same unit