I've assigned @jkczyz as a reviewer!
I'll wait for their review and will help manage the review process.
Once they submit their review, I'll check if a second reviewer would be helpful.

I've completed a thorough review of every hunk in this PR diff, examining:

1. PaymentQueue idempotency (payment_queue.rs): The duplicate intercept_id check correctly prevents re-adding replayed HTLCs across all state machine arms that call add_htlc. The debug_assert_eq on payment_hash is appropriate validation.

2. remove_terminal_channel_state (service.rs): Correctly guards removal behind PaymentForwarded state check via get_channel_id() and channel_id match. Cleans up all three inner maps consistently.

3. channel_closed (service.rs): Properly cleans up both outer maps (peer_by_intercept_scid, peer_by_channel_id) and inner state. Uses self.persist() rather than persist_peer_state(), which is the right choice since it also handles pruning empty peer state.

4. Sync wrapper (service.rs): Follows the established pattern used by all other sync wrappers in LSPS2ServiceHandler.

5. Documentation (manager.rs): Correctly adds ChannelClosed to the event forwarding list with proper doc links.

6. Tests: Both new tests (replayed_intercepted_htlc_after_persist_is_idempotent and removes_terminal_state_for_closed_channel) correctly exercise the key scenarios including serialize/deserialize round-trip and idempotent second removal.

No new issues found. The previously flagged Vec vs HashSet performance concern (inline comment on service.rs:1841) remains the only finding.