Table Maintenance: Cleanup/Prune Safety and Concurrency Caveats Undocumented (All SDKs)

[oqoqo-bot]

Documentation Gap

The LanceDB docs do not document cleanup/prune safety controls, concurrency caveats, or the delete_unverified option across any SDK (Python, TypeScript, Rust).

Description

Table cleanup/prune safety controls, concurrency behavior, and the delete_unverified option are undocumented across all SDKs.

• All SDKs expose table.optimize() with cleanup/prune and a delete_unverified option, but the docs don't explain these — the 7-day default retention protects in-progress transactions, yet this rationale is only in code docstrings
• Concurrency behavior is undocumented: concurrent operations typically succeed due to version isolation, but cleanup itself may fail or be conservative when other operations are active
• Users on Discord are asking whether aggressive cleanup (e.g., cleanupOldVersions(0)) risks data loss — cleanup failures don't risk data loss and retries are safe, but users have no way to know this

How to Validate

• Confirm versioning docs do NOT explain cleanup safety, concurrency behavior, or the delete_unverified option.
• Confirm lance.mdx does NOT document how to run table.optimize() or cleanup manually.
• Confirm delete_unverified and 7-day default behavior are NOT explained in user-facing docs (only in code docstrings).
• Review Discord thread confirming users asked these questions because docs were unclear.

Affected Files

• docs/tables/versioning.mdx
• docs/lance.mdx
• rust/lancedb/src/table.rs

---

Created by Oqoqo

[prrao87]

Covered in #154. Closing this.