[Devbook drift] .devbook/domain/sessions/features.md#honest-partial-answers #879

Scheduled Devbook verify run, 2026-09-30 (devbook:verify-change, kind feature).

Evidence: Sub-feature Name the source that could not be read counts "never installed" as unreadable and named; the passing test AgentSessionSourceTests.An_agent_that_was_never_installed_is_not_an_unreadable_source asserts catalog.Unreadable is empty when an agent's folder is absent. Say how much was left out matches: per-agent cap of 100 with both numbers in the subtitle (AgentSessionLimits.PerAgent, SessionsPaneTests.A_capped_list_names_both_numbers, An_uncapped_list_does_not_hedge_about_being_complete), uncapped horizon readings (A_reading_since_a_horizon_is_every_claude_session_inside_it_and_not_the_newest_hundred, ..._copilot_...), 12-week retention for replicated records (FileReplicatedSessionStore, ReplicatedSessionLimits.History); an unreadable source is named with a Retry (An_unreadable_source_is_named_above_the_rows_that_did_arrive)

Detail

• The chapter says: when one agent's records cannot be read, "never installed, or not readable by this user", the unreadable one is named.
• The code says an absent (never-installed) agent is not unreadable. The passing test tests/Backlog.Infrastructure.Sessions.UnitTests/AgentSessionSourceTests.cs An_agent_that_was_never_installed_is_not_an_unreadable_source asserts Assert.Empty(catalog.Unreadable). Only permission or read failures are named (SessionsPaneTests.An_unreadable_source_is_named_above_the_rows_that_did_arrive).
• Question: is "never installed" an unreadable source to name, as the chapter says, or an ordinary empty source, as the code and test say?

Action: a person decides which side is the model: the chapter (then the code is a defect, for apply-change) or the code (then the chapter is stale, for capture-specs). The question is under Detail above.

Read-only: nothing was run or written. The chapter carries no draft or proposed status, and no open pull request or change names it.