{
  "summary": "Read-only integration-test census at the supplied fixed worktree: 219 Rust files across six crates/*/tests roots. Found 64 Brain connection/handoff retry loops, 20 connected/reconnecting Brain RPC readiness or delivery loops, and 3 finite Brain event-completion waits. Separately classified 2 finite Brain absence-check drains and adjacent non-Brain polling/workload loops as exclusions, not defects. No edits, scripts, tests, builds, linters, formatters, or validation commands executed.",
  "files": [
    {
      "path": "crates/spt-daemon/tests",
      "description": "84 Rust files (79 root tests, support/serving.rs, four fixture files); contains 61 connection/handoff retry sites, 19 RPC readiness/delivery sites, three event-completion waits, and two absence-check drains."
    },
    {
      "path": "crates/spt/tests",
      "description": "115 Rust files (109 root tests, four fixtures, two common modules); contains three Brain cold-connect helpers and one indirect cold-connect+sessions readiness loop."
    },
    {
      "path": "crates/spt-term/tests",
      "description": "10 Rust files; terminal/session stream polling and workloads excluded from Brain IPC census."
    },
    {
      "path": "crates/spt-store/tests",
      "description": "Seven Rust files; store/process/file operations and fixed workloads, no qualifying Brain IPC iteration bound."
    },
    {
      "path": "crates/spt-msg/tests",
      "description": "One Rust file; ReadyAgent recv/drain_spool windows excluded as relay/spool polling."
    },
    {
      "path": "crates/xtask/tests",
      "description": "Two Rust files; no qualifying Brain IPC iteration bound."
    },
    {
      "path": "crates/spt-daemon/src/inject.rs",
      "description": "Read only to trace integration test drain_idle_spool through drain_spool_offering and try_spt_hosted_inject to Brain cold_start/inject_endpoint."
    },
    {
      "path": "crates/spt-daemon/src/daemon.rs",
      "description": "Read is_running definition to establish seedmap ping boundary, not Brain RPC."
    },
    {
      "path": "crates/spt-daemon/src/tunnelhub.rs",
      "description": "Read tunnel_recv definition to distinguish dedicated tunnel transport from Brain methods."
    },
    {
      "path": "crates/spt/src/cli.rs",
      "description": "Located shell tunnel recv dispatch to tunnel_recv at line 23684."
    },
    {
      "path": "crates/spt/src/api/reporting.rs",
      "description": "Located API tunnel recv dispatch to tunnel_recv at line 753."
    }
  ],
  "architecture": "Integration rigs use Brain::cold_start/cold_start_pump or handoff to acquire broker IPC clients, then poll Brain net_streams/net_status/sessions or consume BrokerEvents. Two readiness/delivery loops reconnect through helpers. Raw LocalSocketTransport+Role::Brain fixtures, seed-control ping, tunnelhub requests, in-process NetHost/Broker state, relay/spool polling and file/process state are separate boundaries. No LSP tool was available in the attached tool inventory; all references are text-search/read evidence, not claimed LSP references.",
  "report": "# Scope and method\n\nFixed tree: `.worktrees/318-h3-infra-register`, supplied snapshot `cd0e4216ef034955e4ea5fa38ede62f11acd1075`. Did not inspect moving main or agent/context files.\n\nPopulation: **219 `.rs` files** under all six existing integration roots: `crates/spt/tests/**/*.rs` (115), `crates/spt-daemon/tests/**/*.rs` (84), `crates/spt-term/tests/**/*.rs` (10), `crates/spt-store/tests/**/*.rs` (7), `crates/spt-msg/tests/**/*.rs` (1), `crates/xtask/tests/**/*.rs` (2). Includes nested fixtures/common/support. Source inline test modules are outside this assignment.\n\nSearch predicates covered ordinary `for ... in ...`, all range spellings (inclusive, typed suffix, variable/constant upper bounds, nonzero lower bounds), `while`, counters (`attempts`, `retries`, `remaining`, `tries`, `poll`, comparison/increment guards), `.take(...)`, `repeat`, `repeat_with`, `successors`, retry helpers, and non-range iterator/array loops. Numeric candidate search was paged through all 127 matching files; supplemental for/while/helper searches classified nonnumeric bounds. Candidate bodies and containing helpers/tests were inspected. `take()` on Option handles, string repeat, byte generation, parsing loops and data/assertion iteration were distinguished from polling. No additional qualifying counter-controlled or repeat/take retry loop was identified in these roots.\n\nCounts below count **lexical loop sites**, not helper invocations or test executions. A bound is an attempt count, not a wall-clock duration; cadence is the explicit sleep, not a claim about actual elapsed time. No failure-frequency, global blindness, or defect claim is made.\n\n# A. Brain connection/handoff retry sites — 64 loops\n\nEvery row below is a **connection acquisition/reacquisition retry**, not an already-connected Brain readiness poll. `cold` = `Brain::cold_start`; `pump` = `Brain::cold_start_pump`; `handoff` = `Brain::handoff`. All have explicit **10 ms sleep after an unsuccessful attempt**. Grouped rows list distinct loop sites separately. Bounds are literal range spelling; counts are their exclusive-upper-bound counts.\n\n| Repo-relative file | Loop line → containing function/test | Literal bound / count | Operation |\n|---|---|---|---|\n| crates/spt-daemon/tests/access.rs | 70 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/attach.rs | 80 → connect_retry; 92 → connect_retry_pump | each 0..300 / 300 | cold; pump |\n| crates/spt-daemon/tests/attach_idempotent_replay.rs | 104 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/attach_resize_capture.rs | 144 → pump_conn | 0..300 / 300 | pump |\n| crates/spt-daemon/tests/attach_resize_repaint.rs | 111 → pump_conn | 0..300 / 300 | pump |\n| crates/spt-daemon/tests/brain_read_deadline.rs | 87 → connect_whole; 97 → connect_pump | each 0..300 / 300 | cold; pump |\n| crates/spt-daemon/tests/brain_swap.rs | 69 → cold_retry; 84 → handoff_relaunch | each 0..300 / 300 | cold; handoff |\n| crates/spt-daemon/tests/budget.rs | 206 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/control_stamp_lifetime.rs | 87 → connect_brain | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/controller_lease.rs | 79 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/daemon_e2e.rs | 153 → cold_retry; 163 → handoff_retry | each 0..300 / 300 | cold; handoff |\n| crates/spt-daemon/tests/digest_cross_node.rs | 87 → connect_retry; 106 → connect_retry_pump | each 0..300 / 300 | cold; pump |\n| crates/spt-daemon/tests/dispatch.rs | 82 → connect_retry; 102 → connect_retry_pump | each 0..300 / 300 | cold; pump |\n| crates/spt-daemon/tests/driven_by_selfheal.rs | 120 → connect_brain | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/endpoint_lifecycle.rs | 91 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/endpoint_survival.rs | 128 → daemon_restart_reruns_a_previously_online_orphaned_endpoint | 0..200 / 200 | cold().is_ok() |\n| crates/spt-daemon/tests/endpoint_survival.rs | 193 → daemon_restart_resumes_a_ready_agent_controllable_orphan | 0..200 / 200 | cold().is_ok() |\n| crates/spt-daemon/tests/exit_every_reap.rs | 79 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/handoff.rs | 59 → cold_retry; 69 → handoff_retry | each 0..300 / 300 | cold; handoff |\n| crates/spt-daemon/tests/idempotent.rs | 65 → cold_retry; 75 → handoff_retry | each 0..300 / 300 | cold; handoff |\n| crates/spt-daemon/tests/inject_control_wedge.rs | 582 → connect_retry; 1029 → spawn_session_with; 1301 → spawn_xlate_session | each 0..300 / 300 | cold; pump; pump |\n| crates/spt-daemon/tests/input_ack_deadlock.rs | 187 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/mesh.rs | 94 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/mesh_recovery.rs | 81 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/netbroker.rs | 83 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/netstream.rs | 58 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/notifsync.rs | 63 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/pairjoin.rs | 73 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/presence.rs | 56 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/propagate.rs | 76 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/pump.rs | 120 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/redispatch.rs | 77 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/redispatch_stall.rs | 92 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/registry_lifecycle.rs | 77 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/render_lifecycle.rs | 63 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/replicate.rs | 64 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/resize_geometry_epoch.rs | 90 → pump_conn | 0..300 / 300 | pump |\n| crates/spt-daemon/tests/resize_presentation_barrier.rs | 94 → pump_conn | 0..300 / 300 | pump |\n| crates/spt-daemon/tests/restart_replay_lifetime.rs | 181 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/resume.rs | 53 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/spawn_truth.rs | 54 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/sync.rs | 89 → connect_retry | 0..300 / 300 | cold; separate from repaired stream wait |\n| crates/spt-daemon/tests/transport_death_eof.rs | 220 → connect_retry | 0..300 / 300 | cold |\n| crates/spt-daemon/tests/twohost.rs | 443 → connect_retry; 464 → connect_retry_pump | each 0..300 / 300 | cold; pump |\n| crates/spt-daemon/tests/twohost.rs | 1604 → two_host_ladder_role_b (ack_brain block) | 0..300 / 300 | pump |\n| crates/spt-daemon/tests/twohost.rs | 2921 → two_host_ladder_role_a (redeem_brain block); 3085 → same test (dig_a block) | each 0..300 / 300 | pump; pump |\n| crates/spt-daemon/tests/wan_reply_bound.rs | 98 → connect_retry; 117 → connect_retry_pump | each 0..300 / 300 | cold; pump |\n| crates/spt-daemon/tests/wanmsg.rs | 77 → connect_retry | 0..300 / 300 | cold |\n| crates/spt/tests/brain_survive.rs | 130 → connect_retry | 0..400 / 400 | cold |\n| crates/spt/tests/coordinator_image_e2e.rs | 49 → connect_retry | 0..400 / 400 | cold |\n| crates/spt/tests/twohost_cli.rs | 540 → connect_retry | 0..300 / 300 | cold |\n\nThese 64 comprise 60 cold/pump acquisitions and four handoff retries. `twohost` files remain part of the lexical census even though their tests are environment-gated.\n\n# B. Brain RPC readiness/delivery loops — 20 sites\n\nUnless marked reconnecting, the loop calls methods on an already-connected Brain. All stream waits issue a new broker RPC each iteration; their observation of network stream state does **not** make them in-memory NetHost polling.\n\n| Repo-relative file:loop line | Function/test | Bound / count | Explicit sleep | IPC / condition |\n|---|---|---|---|---|\n| crates/spt-daemon/tests/access.rs:80 | wait_for_stream | 0..400 / 400 | 5 ms | net_streams; first row |\n| crates/spt-daemon/tests/attach.rs:140 | wait_for_stream | 0..400 / 400 | 5 ms | net_streams; peer-initiated row not in served |\n| crates/spt-daemon/tests/controller_lease.rs:165 | wait_for_new_stream | 0..400 / 400 | 5 ms | net_streams; peer-initiated row not in seen |\n| crates/spt-daemon/tests/endpoint_lifecycle.rs:606 | raw_viewport_close_frees_the_controller_full_chain_across_broker_restart | 0..400 / 400 | 5 ms | target.net_streams; inbound attach row |\n| crates/spt-daemon/tests/exit_every_reap.rs:193 | observe_a_reap | 0..400 / 400 | 5 ms | table.net_streams; inbound attach row |\n| crates/spt-daemon/tests/inject_control_wedge.rs:610 | wait_for_stream | 0..400 / 400 | 5 ms | net_streams; peer-initiated row |\n| crates/spt-daemon/tests/input_ack_deadlock.rs:199 | wait_for_stream | 0..400 / 400 | 5 ms | net_streams; peer-initiated row |\n| crates/spt-daemon/tests/input_ack_deadlock.rs:213 | wait_for_latest_stream | 0..400 / 400 | 5 ms | net_streams; latest peer-initiated row |\n| crates/spt-daemon/tests/netstream.rs:84 | wait_for_stream | 0..400 / 400 | 5 ms | net_streams; first row |\n| crates/spt-daemon/tests/notifsync.rs:77 | wait_for_stream_except | 0..400 / 400 | 25 ms | net_streams; row not in known |\n| crates/spt-daemon/tests/propagate.rs:89 | wait_for_stream_except | 0..400 / 400 | 25 ms | net_streams; row not in skip |\n| crates/spt-daemon/tests/render_lifecycle.rs:141 | final_output_frame_precedes_exit_through_the_production_path | 0..400 / 400 | 5 ms | table.net_streams; inbound attach row |\n| crates/spt-daemon/tests/replicate.rs:75 | wait_for_stream | 0..400 / 400 | 5 ms | net_streams; first row |\n| crates/spt-daemon/tests/wanmsg.rs:89 | wait_for_stream | 0..400 / 400 | 5 ms | net_streams; first row |\n| crates/spt-daemon/tests/netbroker.rs:140 | brain_dials_second_broker_over_loopback | 0..400 / 400 | 5 ms | brain_b.net_status().conn_count == 1 |\n| crates/spt-daemon/tests/presence.rs:167 | presence_history_replays_from_zero_for_a_late_subscriber | 0..400 / 400 | 5 ms | probe.net_status().conn_count == 1 |\n| crates/spt-daemon/tests/exit_every_reap.rs:273 | observe_a_reap | 0..200 / 200 | 25 ms | table.sessions; target session row disappears |\n| crates/spt-daemon/tests/control_stamp_lifetime.rs:265 | converge_stamps_on_sessions_poll_after_late_bind | 0..50 / 50 | 50 ms after sessions RPC, before file check | RECONNECTING: connect_brain then sessions; controlled stamp becomes true |\n| crates/spt/tests/run_no_dup_session_e2e.rs:212 | run_over_live_endpoint_never_mints_a_second_session | 0..60 / 60 | 200 ms | RECONNECTING: broker_session_ids_for; nonempty endpoint sessions |\n| crates/spt-daemon/tests/inject_control_wedge.rs:3237 | parked_message_to_already_idle_endpoint_is_delivered_without_a_poke | 0..40 / 40 | 100 ms | RECONNECTING indirect delivery retry: drain_idle_spool > 0; cold_start + inject_endpoint |\n\nHelper chains (locations are explicit so the caller cannot be misclassified as a file-only poll):\n\n* `control_stamp_lifetime.rs:265` → `connect_brain` definition `:86`, nested retry loop `:87`, cold_start `:88` → sessions `:267` → sleep `:268` → info-file condition `:269`. The inner connection retry is independently counted in A.\n* `run_no_dup_session_e2e.rs:212` → call `:213` → `broker_session_ids_for` definition `:115` → cold_start `:116` → sessions `:123`. This is one outer readiness loop whose helper creates a new Brain each pass; not a standalone connection retry loop.\n* `inject_control_wedge.rs:3237` → `drain_idle_spool` call `:3238` → `crates/spt-daemon/src/inject.rs:128` → `drain_spool_offering` definition `:154` → `try_spt_hosted_inject` call inside claimed-row iteration `:190` → definition `:75` → `Brain::cold_start` `:89`, `brain.inject_endpoint` `:90`. It retries offering a real pending row until delivered; the adjacent log/spool-only waits at test lines 1972 and 3410 are excluded below.\n\n# C. Finite Brain event-completion waits — 3 sites\n\nKeep distinct from repeated RPC readiness queries: these drain a connected carrier awaiting a particular reply/output condition, not repeated cold connects or workload generation.\n\n| Repo-relative file:loop line | Function/test | Bound / count | Sleep / read deadline | Operation |\n|---|---|---|---|---|\n| crates/spt-daemon/tests/attach.rs:1191 | read_outcome | 0..2000 / 2000 | no sleep; deadline = brain.call_deadline() once at :1190 | read_event_until; await BrokerEvent::Subscribed |\n| crates/spt-daemon/tests/controller_lease.rs:125 | read_outcome | 0..2000 / 2000 | no sleep; deadline = brain.call_deadline() once at :124 | read_event_until; await BrokerEvent::Subscribed |\n| crates/spt-daemon/tests/idempotent.rs:215 | pty_writes_are_exactly_once_across_brain_crashes | 0..2000 / 2000 | no sleep; deadline = brain.call_deadline() once at :214 | read_event_until; drain until all TOTAL=8 markers observed; break on read error |\n\n# D. Borderline finite Brain absence checks — 2, EXCLUDED from readiness residuals\n\nThese still call Brain directly and must not disappear from the census, but they assert absence rather than wait for readiness:\n\n| Repo-relative file:loop line | Function/test | Bound | Cadence / operation | Classification |\n|---|---|---|---|---|\n| crates/spt-daemon/tests/attach.rs:1308 | same_origin_re_subscribe_does_not_displace | 0..50 / 50 | no sleep; first.read_event_until(now + 20 ms), breaks on error or Displaced | finite assertion drain: no Displaced after same-origin re-take |\n| crates/spt-daemon/tests/pump.rs:959 | pump_w2_presence_reaches_only_the_subscribing_carrier | 0..5 / 5 | no sleep; nosub.read_event_until(now + 200 ms), breaks on error or NetPresence | finite assertion drain: no presence delivered to non-subscriber |\n\n# E. Adjacent excluded boundaries (not defects)\n\nThe following counts enumerate the cited lexical sites, not all ordinary iteration in the repository. Paths abbreviated `D/` below expand to `crates/spt-daemon/tests/`, `C/` to `crates/spt/tests/`.\n\n## Raw transport connection acquisition — 16 sites, EXCLUDED\n\nThese retry `LocalSocketTransport::connect`, often then `send_hello(..., Role::Brain)`, but do not call Brain methods:\n\n* `D/brain_decouple.rs:124`, `D/brain_resume_conn_deadlock.rs:103`, `D/conn_blackhole_lifecycle.rs:121`, `D/false_promote.rs:142`, `D/wake_single_flight.rs:95` — `connect`, each `0..200`.\n* `D/inject_control_wedge.rs:296,395` — direct transport connect retries, each `0..200`.\n* `D/broker.rs:80,209,332,441,556,694,1012` — seven transport connect retries, each `0..200`; :694 is helper `connect_brain`, whose name alone is not proof of Brain abstraction usage.\n* `C/resume_no_control_steal_e2e.rs:216`, `C/trial_drain_drive_e2e.rs:232` — `connect`, each `0..300`.\n\n## Raw protocol/frame polling — 6 sites, EXCLUDED\n\n`D/broker.rs:476` (`0..200`, spawned frame), `:491` (`0..100`, send subscribe, 20 ms sleep), nested `:503` (`0..30`, exit/error frames), `:618` (`0..400`, endpoint-id output), `:811` (`0..4000`, applied/output), `:934` (`0..2000`, marker output). These use raw Stream/read_frame/send rather than Brain. The nested subscribe loop is genuine protocol polling, not workload repetition, but outside the requested Brain-method residual set.\n\n## Seed-control readiness — 23 sites, EXCLUDED\n\nAll `0..400` around `spt_daemon::is_running()`. Definition is `crates/spt-daemon/src/daemon.rs:603-605`: `seedmap::ping(&seed_socket_name()).is_ok()`, not Brain::cold_start/net_status.\n\n`C/bind_adapter_profile_persist_e2e.rs:44`; `boundary_ready_strand_e2e.rs:33`; `contract_e2e.rs:51`; `create_bind_rest_active_e2e.rs:41`; `drive_e2e.rs:70`; `gateway_e2e.rs:54`; `gateway_owner_shell_e2e.rs:112`; `io_state_payload_e2e.rs:61`; `listen_seed_retry_e2e.rs:42`; `live_bind_firsthost_e2e.rs:40`; `live_firsthost_e2e.rs:44`; `live_resolve_e2e.rs:42`; `notify_shell_e2e.rs:35`; `psyche_sid_custody_e2e.rs:50`; `shell_actgate_e2e.rs:36`; `shell_e2e.rs:37`; `shell_hints_e2e.rs:49`; `shell_perch_dir_e2e.rs:38`; `shell_relink_force_e2e.rs:43`; `shell_sleepwake_e2e.rs:39`; `shell_stale_online_e2e.rs:43`; `shortform_dispatch_e2e.rs:44`; `tunnel_e2e.rs:74`.\n\n## Dedicated tunnel transport via CLI — 5 sites, EXCLUDED\n\n* `C/tunnel_e2e.rs:190` shell_recv_until and `:204` owner_recv_until: `0..400`, 10 ms sleep, accumulate CLI recv bytes.\n* `C/gateway_owner_shell_e2e.rs:272` shell_recv_until and `:287` owner_recv_until: `0..400`, 10 ms sleep, same transport.\n* `C/tunnel_e2e.rs:273`: `0..30`, 5 ms sleep, repeated recv asserting absence of stale-generation bytes.\n\nChain: shell CLI `crates/spt/src/cli.rs:23684` or API `crates/spt/src/api/reporting.rs:753` → `spt_daemon::tunnel_recv` → `crates/spt-daemon/src/tunnelhub.rs:440-451` opens its own transport and exchanges KIND_TUNNEL_RECV. Server at `tunnelhub.rs:352-359` drains broker-owned net state directly. No Brain client method in these recv loops.\n\n## In-memory NetHost and Broker state — 6 sites, EXCLUDED\n\n* `D/reseed.rs:81`, `rosterprop.rs:109`, `seedproofx.rs:45`: three `wait_until` helpers, each `0..400`, 5 ms sleep. Predicates inspect NetHost conn counts, in-memory adoption collector, or roster store. Explicit helper uses: reseed :136,:147,:180,:216; rosterprop :178; seedproofx :84,:114,:130,:146. No Brain IPC in these helper predicates.\n* `D/brain_swap.rs:186`, `daemon_e2e.rs:373`, `handoff.rs:192`: three `0..200` waits on `broker.session_count()` directly, 10 ms sleep; not `Brain::sessions`.\n\n## File/store/process observation — 28 sites, EXCLUDED\n\nThese loops' bodies read local facts rather than making Brain RPCs, even where another thread runs a Brain:\n\n* `D/control_stamp_lifetime.rs:191`.\n* `D/dispatch.rs:357,440,856,875,1190` — notif/spool store reads.\n* `D/endpoint_lifecycle.rs:150,258,349,634,660,855,883` — process liveness/relay ownership pin or info stamps. In particular :855/:883 do not become Brain RPC loops because an independently stop-flag-driven sessions poller exists at :871.\n* `D/inject_control_wedge.rs:1972,3410` — binary stdin log / spool pending count (each `0..60`, 100 ms).\n* `D/pump.rs:634,644,855,864` — heartbeat file observation, not net_status RPC.\n* `D/restart_replay_lifetime.rs:426,574` — info controlled stamp.\n* `D/spawn_truth.rs:358,374,406,462` — process descendants/existence/liveness.\n* `D/transport_death_eof.rs:490` — info controlled stamp.\n* `D/shell_kill_auth.rs:87` — child try_wait.\n* `C/job_escape_e2e.rs:491` — child/control process observation.\n\n## Relay, terminal stream, listener-close polling — 9 sites, EXCLUDED\n\n* Relay/spool/absence: `D/endpoint_lifecycle.rs:395` (`0..100`, sibling_relay.pump_once); `D/notif_quiet_delivery.rs:130` (`0..50`, listener.check_message absence assertion); `crates/spt-msg/tests/killer_quickstart.rs:35` (`0..windows`, ReadyAgent.recv plus drain_spool, no explicit sleep in helper).\n* Terminal-only: `crates/spt-term/tests/dsr.rs:36`, `inject.rs:37,104`, `stream.rs:56,95` — output buffers, Session.try_wait, or terminal stream.recv_timeout; five sites, no Brain IPC.\n* `D/lan_bootstrap_e2e.rs:351` — `0..50` TcpStream listener-close observation, not Brain.\n\n## Fixed-count Brain-adjacent workload/assertion repetitions — 32 explicitly identified sites, EXCLUDED\n\nThese iterate work/cases or assert every item; they are not readiness retries. This named subset includes the easy-to-misread variable/constant-bound cases:\n\n* `D/brain_resume_conn_deadlock.rs:195` (`0..n_sessions`, create flood sessions), `:311` (`0..N`, N=3, create fixture sessions).\n* `D/budget.rs:82` (`0..n`, create seats).\n* `D/controller_lease.rs:578` (`0..300u64`, send flood).\n* `D/daemon_e2e.rs:270` (`1..=RESTARTS`, RESTARTS=10), `D/handoff.rs:111` (`1..=RESTARTS`, RESTARTS=12): restart workloads. Their inner handoff retry helpers remain included in A.\n* `D/idempotent.rs:124` (`0..total`, issue effects), `:189` (`gauntlet`, crash cases), `:241,:256` (`0..TOTAL`, TOTAL=8, per-op assertions). The separate event drain :215 is included in C.\n* `D/inject_control_wedge.rs:1497` (`0..4`, input/flush probes), `:1614` (`0..6`, send secret keystrokes), `:2057` (`0..3`, fault three generations), `:3322` (`0..N`, N=4, distinct envelopes).\n* `D/input_ack_deadlock.rs:417` (`1..=FLOOD_N`, FLOOD_N=64, send load).\n* `D/netstream.rs:161` (`0..KILL_AT`, KILL_AT=5), `:184` (`KILL_AT..TOTAL`, TOTAL=20), `:206` (`0..TOTAL`, construct expected bytes), `:238` (`0..CRASH_AFTER`, CRASH_AFTER=3), `:262` (`0..TOTAL`, TOTAL=8, re-drive sends), `:287` (`0..=TOTAL`, journal assertions), `:301` (`0..TOTAL`, output assertions).\n* `D/redispatch.rs:453` (`0..3u32`, marker workload across dispatch cycles), `:627` (`0..40u64`, flood).\n* `D/redispatch_stall.rs:351` (`0..FLOOD_PRE`, 48), `:372` (`0..FLOOD_POST`, 48), `:434` (`0..600u64`, continuous forwarder tick workload with 100 ms sleep), `:456` (`0..FLOOD_PRE as u64`, 48), `:769` (`0..3u32`, marker workload). The forwarder exits on RPC failure; it does not retry until a readiness condition.\n* `D/wanmsg.rs:335` (`0..3u64`, distinct messages).\n* `D/wan_reply_bound.rs:289` (`[\"fork\", \"redeem\", \"answer\", \"knock\", \"presence\"]`, separate verb cases).\n* `C/resume_no_control_steal_e2e.rs:382` (`0..N_SESSIONS`, N_SESSIONS=3, spawn ticker fixtures).\n\nOther ordinary parsing, frame decoding, fixtures, byte generation (`repeat`, `cycle().take`), HTTP test matrices, store workload loops and assertion iteration are likewise not polling residuals. Notably `net_worker_starve.rs:142,:205` (`0..K`, K=12) are NetHost dial workloads, not Brain retries.\n\n# F. Positive controls, deliberately NOT residual\n\n`crates/spt-daemon/tests/sync.rs:98` `wait_for_stream_except`, loop at :107, now checks `started.elapsed() >= Duration::from_secs(60)` at :108 before `brain.net_streams()` :113. Cadence :122 is `25 ms.min(remaining budget)`; IPC error is propagated through the result. `iterations` is telemetry, not a loop bound. This is the landed IR-95 stream-wait remedy positive control in the supplied tree; do **not** re-register it as a fixed-iteration residual. Its unrelated cold-connect helper at :89 remains a distinct category-A site.\n\nOther wall-deadline convergence helpers (`mesh`, `mesh_recovery`, `pump`, `registry_lifecycle`, role rig_wait, coordinator image wait, etc.) are not converted into fixed-count entries merely because they contain an attempts counter or call a separately count-bounded acquisition helper. Likewise the raw-frame `post < 8` guards in `brain_decouple.rs:591` and `conn_blackhole_lifecycle.rs:374` also have explicit wall deadlines and are stream seeding, not Brain-method readiness polls.\n\nNo remedies proposed and no code changed. Main owns documentation edits, landed-commit/PR wording and CI receipts."
}