follow up from #59591
It would be nice if optimizations are transparent, #59591 adds a performance warning if the transform might suffer from expensive lookups. This could be enhanced to tell the reason and potentially how to improve the configuration for better runtime.
