Temporary files should not be visible to OS-level backup software
Web apps may create temporary files that should not be backed up via OS-level backup software. Some web apps may create temporary files that are many GBs in size. Backing up these files consumes disk io, network io and storage resources.
One way of solving this that I had been thinking about would be to add a policy to storage buckets that lets you mark certain buckets as not needing to be backed up, as this also seems useful for certain usage for other storage APIs. I.e. pure cached data would not need to be backed up as the website could easily re-download it. That would then apply to files in OPFS in those specific buckets. Or alternatively chrome could have some heuristics based on expiration time and persistence of a storage bucket. I think being able to exclude data from backups is definitely a gap in the current API surface.
Separately I wonder if more explicit support for temporary files (that might not even need to live in a "file system") would make sense. I.e. just being able to create a file handle/access handle without ever having to create the file behind it (and as such, the website wouldn't be able to create references to the file again later, signalling to chrome that it is more of a scratch space/session scoped in lifetime). But not clear that that would have significant benefits over being able to specify this at the bucket level.