I found a solution to get up-to-date data using the gcloud storage Python library, and it works effectively in "real time."

My guess is that using this approach, the gcloud storage library directly interacts with the Google Cloud Storage API instead of relying on file system synchronization through mounted volumes. This eliminates any potential caching or synchronization lag and ensures that the latest version of the data is always retrieved. I didn't expect such delays.

I'm keeping the post open because the question about delays and consistency when using mounted volumes in Cloud Run still stands. Not having high speed/consistency somehow kills some of the beauty of using the mounted volumes as local paths.