rrt/artifacts/exports/rt3-1.06/runtime-effect-kind8-post-reload-candidates.md

104 lines
5.7 KiB
Markdown

# Runtime Effect Kind-8 Post-Reload Candidate Branches
This note records the currently grounded late bringup branches that remain between ordinary
runtime-effect reload and the final one-shot kind-`8` service under
`world_entry_transition_and_runtime_bringup` `0x00443a50`.
## Ordering Boundary
- `world_load_saved_runtime_state_bundle` `0x00446d40`
- directly restores event runtime records through
`scenario_event_collection_refresh_runtime_records_from_packed_state` `0x00433130`
- does **not** directly call
`scenario_event_collection_service_runtime_effect_records_for_trigger_kind` `0x00432f40`
- `world_entry_transition_and_runtime_bringup` `0x00443a50`
- owns both the reload phase and the later final kind-`8` service
- clears shell-profile latch `[0x006cec7c+0x97]` after that final kind-`8` call
## Tier 1: Direct Live-Event Mutator
Grounded from the current `0x00443a50` note and subgraph artifacts:
- `0x00442c30`
- `shell_apply_scenario_name_specific_post_load_world_and_object_fixups`
- explicit live event/object mutation branch
- already known to rewrite some `[event+0x7ef]` values on scenario-specific cases
## Tier 2: Indirect Candidate / World-State Rebuild Owners
These late owners still matter because they reshape candidate-side or world-side state before the
final kind-`8` service runs, but they do not currently read as direct runtime-event-row
materializers.
- `0x004354a0`
- `world_publish_shell_controller_progress_scalar_from_year_thresholds_or_selector_overrides`
- late shell-controller progress/state publish in the reactivation tail
- `0x00435603`
- `scenario_state_ensure_derived_year_threshold_band`
- late guard over the derived year-threshold band
- `0x00412c10`
- `structure_candidate_collection_refresh_named_availability_latch_0x7ac_and_tail_into_cargo_economy_filter_refresh`
- rerun after late candidate-side availability restoration/preseed
- `0x00412bd0`
- `structure_candidate_collection_rebuild_cargo_membership_tables_counts_and_scaled_rate_bands_for_all_live_records`
- rerun in the same late candidate rebuild band
- `0x00434130`
- `scenario_state_refresh_year_gap_scalar_from_current_year_minus_1850_into_0x4ca2`
- year-gap derived scalar refresh
- `0x00436af0`
- `scenario_state_refresh_cached_available_locomotive_rating_from_year_baseline_and_live_availability_gates`
- cached locomotive-rating refresh
- `0x00437737`
- `world_preseed_named_candidate_availability_records_from_live_pool`
- late candidate preseed branch called after the second recipe rebuild in the current note
## Weaker / Lower-Priority Candidates
- `0x00433130`
- already identified as the ordinary reload owner, not the late materializer
- `0x00446d40`
- already identified as the loader above `0x00433130`, not the final kind-`8` service owner
## Current Bias
- The known title-fixup branch `0x00442c30` remains plausible in general because it explicitly
mutates live runtime-event rows.
- The checked export
`artifacts/exports/rt3-1.06/compact-event-dispatch-cluster-counts.json`
now keeps the ordinary bundle corpus grounded too: the current `rt3_105/maps` set still has
`318` nondirect dispatch-strip rows across `38 / 41` maps, with `0` recovered trigger kinds and
only `10` add-building grouped rows across the `7` descriptor keys
`Barracks`, `Bauxite Mine`, `FarmGrain`, `Furniture Factory`, `Logging Camp`, `Port01`, and
`Warehouse05`.
- The widened full-corpus cluster side now keeps the broader ordinary nondirect families visible
too:
- `dispatch_signature_condition_cluster_occurrence_counts` now shows the broadest checked family
is `nondirect-ge1e-h0001-0360-0004-0100-0200-p0000-0000-0000-ffff :: [864:4]` at `36`
grouped occurrences across `18` maps
- the next family remains the mixed `[-1:4]` cluster at `27` grouped occurrences across `14`
maps
- `dispatch_signature_condition_cluster_descriptor_keys` keeps the `36`-occurrence cluster on
the variable/status lane (`Economic Status`, `Game Variable 1..4`, `Company Variable 1..4`)
and keeps the `[-1:4]` cluster on that same family plus the checked `Barracks` /
`Bauxite Mine` add-building pair
- That same export now also keeps the carrier maps visible without reopening the full cluster
dump: `Texas Tea.gmp` carries `Port01`, `Louisiana.gmp` carries `Warehouse05`,
`Alternate USA.gmp` carries the repeated `FarmGrain` / `Logging Camp` family, and the mixed
`[-1:4]` cluster is split between `Chicago to New York.gmp` and `Pacific Coastal.gmp`.
- The checked periodic-company trace now mirrors the same late candidate and retag evidence too:
- `near_city_acquisition_nontransport_late_bringup_candidate_branches`
- `near_city_acquisition_nontransport_explicit_trigger_kind_retags`
- `near_city_acquisition_nontransport_top_dispatch_signature_condition_clusters`
- But the checked title-hint and direct comparison notes now narrow that path sharply:
- `artifacts/exports/rt3-1.05/add-building-map-title-hints.json`
shows `5 / 6` shipped add-building carrier maps have some grounded title hit, but only
`1 / 6` shows an adjacent embedded `.gmp` reference plus grounded title and only `1 / 6`
shows a same-stem pair.
- `artifacts/exports/rt3-1.06/runtime-effect-kind8-title-overlap-note.md`
shows that strongest overlap (`Louisiana.gmp -> Dutchlantis`) still does not reproduce the
actual `Add Building Warehouse05` dispatch cluster because `Dutchlantis.gmp` has no current
add-building dispatch rows.
- That shifts the best current recovery question toward the late candidate/world-state rebuild band:
- which Tier 2 branch between `0x00433130` reload and final `0x00432f40(kind 8)` service
enables the ordinary shipped-map mutation-capable rows before the live event service consumes
them.