-- Why a phase is waiting, so a queue never looks like a stall.
--
-- Placement moves from mission launch to PHASE launch, which makes the existing
-- `start_pending_phases` loop the queue: a phase with no admissible node simply
-- stays `pending` and is retried every 10s. That is the whole queue mechanism —
-- no new status, no sweeper, no migration for a `queued` state.
--
-- The risk it creates is the one this codebase keeps paying for: a phase waiting
-- for capacity is indistinguishable from a phase nothing is working on. These two
-- columns are what make the difference visible. `capacity_blocked_since` also
-- bounds the wait, so a fleet that never frees fails the phase with a reason
-- instead of holding it forever.
ALTER TABLE mission_phases
    ADD COLUMN IF NOT EXISTS capacity_blocked_since timestamptz,
    ADD COLUMN IF NOT EXISTS capacity_note          text;