LinkedIn API v2 search pagination: bounding limit and cursor lifetime for a large dataset export

LinkedIn's v2 REST search endpoints return results in pages, with a limit parameter controlling page size and cursor-based paging used to walk large result sets. A planned export needs to pull a large dataset through these endpoints in bounded requests rather than one unbounded call, and the pagination configuration has to be settled before implementation.

The open decision is how to bound the walk. The per-request limit is reportedly capped at a small maximum, and pagination cursors are said to have a limited lifetime, which matters for an export that may pause and resume. It is also unclear whether offset-style start paging remains a supported alternative for these endpoints, or whether it truncates results beyond an undocumented bound.

1. What is the documented maximum page size for v2 search endpoints, and does exceeding it raise an error or silently truncate the response?
2. How long are pagination cursors valid, and what is the expected checkpoint strategy when a large export must resume mid-dataset?
3. Is offset-based paging reliable for large exports, or is cursor-based paging the only supported mechanism?