@pp-mo has some ideas on this

End goal: replace existing Iris memory benchmarking with a tracemalloc approach, hopefully as a basis for almost all SciTools memory benchmarking in future

Brief team discussion Thu 9th afternoon