I think the mode we use those docker images in is typically a single

docker run ... <cmd>

, so generally caches like that wouldn’t survive run over run