Make the Claude Code session import limit configurable

Problem

The Claude Code import flow currently lists at most 50 sessions. I have roughly 6,000 local Claude Code sessions and want to migrate the complete history into Codex Desktop, but sessions beyond that first/recent set cannot be discovered or selected.

A fixed limit is reasonable as a page size, but it should not be the maximum importable history.

Expected behavior

Provide at least one of these options:

• an Import all sessions action
• a configurable import limit
• paginated selection that can continue until every source session has been discovered

Suggested implementation

• Keep 50 as the default discovery page size
• Discover additional sessions lazily with cursor-based pagination
• Import in bounded batches rather than loading every transcript into memory
• Show total progress and allow cancellation/resume
• Use stable source-session identity so retries are idempotent and do not create duplicates
• Preserve source timestamps and project metadata where available (related: #32351)

Why this matters

Large Claude Code histories are common for long-running users. A hidden hard cap makes the migration feature useful only for recent history and prevents adopting Codex as the primary session archive.

Acceptance criteria

• More than 50 Claude Code sessions can be imported without patching the app
• There is no fixed global maximum imposed by the UI
• Importing thousands of sessions remains responsive through batching/pagination
• An interrupted import can resume without duplicating sessions

Environment

• Codex Desktop: 26.803.10989.0
• Platform: Windows 11 x64
• Source history: approximately 6,000 local Claude Code sessions

Start by locating the Claude Code import flow and the current 50-session discovery limit in Codex Desktop. Define how pagination or configurable limits, bounded batching, progress, cancellation/resume, stable source identity, and metadata preservation satisfy the acceptance criteria without imposing a global maximum.