Like i said in an email, those match the corrupted files at ~99,5%, the remaining ones are files whose size > 4Gib.
We are good for s3 bucket deletion.

Latest missing files (24348 items)

there is now a page on the WG wiki to do this: https://wg.softwareheritage.org/index.php?title=Glossary

Thanks.
Missing data reinjected.

Check done (source file : a s3 listing Guillaume provided me in december). 85670 unique sha1s are missing.

Currently independently running a final check of what data has been properly imported from s3 before removing it from the source.

I forgot to update this, sorry.