Efficiencies performing check or sync on b2 bucket

When performing a check or sync with B2, does rclone only incur the transactional overhead of querying checksums versus downloading?

Rclone has to download the directory listings from B2 which contain the sha1sum in them. It doesn’t download the file unless you use the --download file with rclone check

Ok that clarifies it. Given that I am new to managing cloud backups manually, is the functionality sparse whereby check leverages the providers resulting checksum? Is there a chance where the hash in the listings is different than what is on disk at the provider, or is that value a result of the providers return value compared with what you intended to send?

I am just trying to understand the risk associated with each approach so I can be most confident.

Thanks for taking the time to help me!

Rclone computes the hash of the whole file locally and retreives the hash from the cloud provider and compares them.

How the cloud provider calculates the hash varies, but I think in most cases it is calculated when the file is uploaded and stored as metadata with the file. Some cloud providers make a point of telling you that then compare checksums regularly (eg Wasabi) but some don’t. For some cloud providers, rclone sends the checksum as metadata and the cloud provider just stores it - this isn’t common though.

sha1sums at B2 are handled very well and are a useful check on data integrity.