# F63 baseline — files with unbounded ``.fetchall()`` calls that lack
# a ``LIMIT`` in the query or a ``# F63-bounded:`` rationale on/above
# the call line. Grandfathered; shrink-only per F49.
#
# As of 2026-06-06: empty. Every previously-listed call site
# (cc_pair, collection_router, dead_letter, scope_profile_resolver,
# streaming_bronze, integrity, repository, embed/embed, embed/schema,
# facts/store, topology_v2_status, search/vec_index, eval/auto_gold)
# carries a ``# F63-bounded: <rationale>`` comment naming why the
# unbounded fetchall is safe at the scale it will hit:
#
#   * PRAGMA table_info — schema-bounded (rows = column count)
#   * Topology / scope / cc_pair / dead_letter — operator-config-sized
#   * Integrity preflight — startup/operator-invoked, not tick-loop
#   * Streaming bronze replay — operator-invoked rebuild, not tick-loop
#   * Embed worker tick — consumes all candidates by design; per-batch
#     processing upstream bounds memory pressure. Scale risk flagged
#     for future streaming-cursor refactor (see #211).
#   * Eval auto_gold — CLI build tool, not tick-loop
#
# The embed worker fetchalls are the genuine future-paydown target:
# converting them to streaming cursors removes the materialised row
# list at the boundary. F50 still blocks net-new violations from
# accreting to this list.
