Context

In disaggregated storage, a follower node currently must pick up a checkpoint via the standard checkpoint flow before any pages from shared storage can be read. All page access is gated behind a successful checkpoint pickup.

Motivation

When a checkpoint is corrupt, the follower cannot pick it up and therefore loses all access to the underlying pages — even pages that are themselves intact in shared storage. This makes recovery and forensic investigation harder than it needs to be: an engineer has no way to inspect or salvage individual pages when checkpoint metadata is the thing that's broken.

Definition of Done

1. Add a follower-mode path that can fetch and read individual pages from shared storage without requiring a checkpoint pickup.
2. Path must work when the latest checkpoint is corrupt or unreadable (page reads do not depend on checkpoint metadata being valid).
3. Surface the capability through a tool/API suitable for diagnostic and recovery use (e.g. via wt util or the pagelog tooling).