First byte and final byte are different clocks

Readiness and delivery answer different questions

An archived object may need a restore or rehydration step before it can be read. The selected retrieval option affects that readiness step. It does not measure delivery to the recovery destination. [1] [2]

A provider may describe one object or one retrieval request. A full backup can contain many objects, and an account-level limit can constrain the whole job. Do not multiply a per-object availability statement into an assumed bulk-recovery guarantee. [2]

What a complete recovery-time estimate needs

Record the restore volume, object count, retrieval option, destination and a measured sustained-throughput range. Provider limits, concurrency, request overhead and the receiving connection can each constrain the transfer. [1] [4] [5]

State whether downloads start as individual objects become readable or wait until the whole set is ready. Retries, integrity checks and the application recovery procedure are additional work; this site has not benchmarked them.

A bandwidth-only calculation is a transfer model, not a promised recovery time. If the required throughput or workflow evidence is missing, keep final completion time unknown rather than adding an archive delay to an unsupported transfer estimate.

Keep the bill separate from the clock

A low egress price does not imply a fast connection, and an access-time statement does not identify every charge. Storage, requests, retrieval, temporary-copy storage and destination-specific egress remain separate receipt lines.

This model treats the restore as reading the retained source. If a recovery plan also deletes, replaces or changes the tier of that source, evaluate those operations separately.