Job suddenly takes 10 times longer than usual

We weren’t told the BackendStatistics from Complete log of the job log, but they might help little.
Presumably it would show a small upload, but it might be a few new files plus a lot of new metadata.

The database view of new metadata would be later paths in File view showing different MetadataID, however keeping the same BlocksetID for the most part (exceptions being files with content change).

Here’s me in DB Browser for SQLite, filtering for file that changes a lot in both content and metadata:

Original post has equivalent log info. Sorry I missed that. Bad run and one after both show added files.
This likely refutes the theory about sets of long runs being due to no file content having been modified.

I asked for OS earlier. I guess I should also ask for filesystem type whose metadata seemed changed.