ShotGrid find pagination stability with limit and offset under concurrent changes

Goal is to define reliable bulk retrieval semantics for ShotGrid API find operations when result sets exceed a single page.

Documented capability uses limit and offset paging for find requests, with a maximum page size in the order of hundreds to a thousand entities. Pagination relies on offset rather than stable cursor tokens. Administrative settings and deployment type can affect effective limits for attachments and API handling.

Uncertainty remains about result consistency when entities are created or deleted between successive pages, and about the inclusion rules for soft-deleted entities across REST API, Python API and web UI. No uniform documented guarantee is available for ordering stability or visibility of deleted revisions in find versus direct read calls.

What ordering guarantees, if any, are documented for paginated find results for a given filter? Does ShotGrid define expected behavior for offset shifts caused by concurrent entity creation or deletion during a multi-page retrieval? Is there a documented pattern for deterministic bulk access that avoids duplicate or missing entities under concurrent modification?