Cleanup and save state

Scene termination is not quest completion, and quest completion is not world cleanup. First Contact makes every boundary visible so one successful spoken line cannot hide a leaked actor, stale marker, or save-backed old graph.

Five different completion boundaries

Do not collapse these into a generic “end.” In particular, there is no quest-scene-node output called end; the public success name is contact_done.

Exact completed lifecycle

The completed child phase orders actor readiness, scene handoff, and teardown:

child Input
  -> meet objective Active
  -> contact mappin Active
  -> community Activate
  -> CharacterSpawned
  -> setup trigger IsInside
  -> checkpoint
  -> scene start
  -> scene contact_done
  -> meet objective Succeeded
  -> contact mappin Inactive
  -> leave objective Active
  -> cleanup trigger IsOutside
  -> community Deactivate
  -> leave objective Succeeded
  -> child Output

After the child returns, the root succeeds the quest journal state, writes cqa005_completed, and exits. On a later activation, its completion guard can bypass the child rather than spawning the contact again.

The order is intentional:

• CharacterSpawned prevents the scene racing a community entry that has not materialized;
• contact_done is consumed before the meet objective and pin change;
• the leave objective makes delayed cleanup visible to the player;
• IsOutside avoids deactivating the contact at speaking distance;
• deactivation completes before the leave objective and child output;
• the durable completion fact is written only after the child returns.

A failed or interrupted scene must not accidentally reach the same success chain. Lab 5 exercises the ordinary cqa005 completion through contact_done. Active-line interruption, CutDestination, and interrupted recovery require separate tests if your scene uses them.

Scene interruption is not cleanup

The scene’s Default interruption scenario reacts when speaker distance rises above 6 and permits return after it falls below 5. With playInterruptLine: 1 and talkOnReturn: 1, it describes how the conversation may interrupt and resume.

That policy is separate from all of these:

• scnEndNode 3 and named exit contact_done;
• the quest scene node’s unconnected Default INT and Default RET outputs;
• its empty interruptionOperations;
• its unconnected CutDestination;
• the wider IsOutside cleanup condition;
• community Deactivate;
• the persistent completion fact.

Do not route a six-unit conversational interruption directly to world cleanup. The player may return within five units and continue. Conversely, reaching End does not remove the community; the questphase retains it until the leave boundary is crossed.

notAllowedToBeFrozen: 0, reapplyInterruptionOperationsAfterGameLoad: 0, and syncToMusic: 0 are exact scene-node values in the fixture. They are not proof of save/reload or cut semantics. Those behaviors require runtime observation.

State that can outlive an archive edit

Treat these as separate save variables while diagnosing the lab:

Replacing a .scene, .questphase, or archive does not clear these domains. Resetting only cqa005_completed is also insufficient: a save can still retain an active child, journal state, checkpoint, scene, or community from an older payload. See Facts, journals, and saves.

Practical cleanup matrix

Run this matrix on Cyberpunk 2077 Windows GOG 2.31a with WolvenKit 8.19.0, ArchiveXL 1.27.0, RED4ext 1.30.0, and redscript 0.5.31. Record the packed payload inventory, exact save history, installed framework versions, and result for every row.

The Lab 5 test procedure covers the supplied checkpoint. Also test active-scene reload and interruption when adding those behaviors.

The frozen campaign begins from two untouched manual-save originals, then Case 1 creates the three named manual seeds under the unchanged candidate. Cases 3/4/7, 5/6/8, and 9/11 load distinct full-slot clones from their respective seed. Close the game before cloning. Never use an autosave, checkpoint save, or a save carried forward from a failed or differently build when comparing the result.

If a case cannot be executed because the game forbids saving at that moment, record that limitation. Do not convert an untested save point into a passing claim.

Failure routing