Tree-sync instrumentation and a stop-marker follow-up fix, backing the MILAB-6303 investigation (Desktop App on slow / distant networks). Spec: milaboratory/text#190.

What this adds

Full introspection of the tree-sync load via MI_LOG_TREE_STAT, computed by reusing the state machine's existing changed flag (negligible per-resource cost):

• Cross-cycle duplication: new / changed / unchanged resources, wasted downlink bytes, stable-metadata re-streams, BFS fetches spent on unchanged resources.
• Per-path detail: streaming (rounds, resource frames, stop-marker frames, stop-markers that triggered a follow-up, traverse-stopped) and BFS (resources actually requested, not-found).
• What changed each cycle: fields added / removed / value-changed, kv changes, ready flips, lock transitions, errors attached, duplicates resolved, resources marked final.

This measured that 52-76% of every polling cycle re-fetches unchanged resources (trees changed by only 2-9 resources over 3 min) on a heavy project.

Fix

The resourceTree stop-marker follow-up now propagates traverseStopRules, so the retry stops at the same boundaries instead of traversing the stop-marked subtrees unbounded. It also asserts that one follow-up round resolves every stop marker (throws if a deeper stop-marker chain remains, which would mean the retry must become a loop).

Status

Draft. Type-check clean; pl-tree unit tests pass (the one skipped failure needs a live PL_ADDRESS). No end-to-end validation of the stop-rule change under a deep tree yet; the assertion is the tripwire for that.

Greptile Summary

Adds detailed tree-synchronization introspection and changes how streaming stop markers and KV mutations are reconciled.

• ResourceUpdateStat / TreeLoadingStat — counters describing fetched resources, duplicate traffic, mutation categories, streaming frames, BFS requests, and stop-marker follow-ups; this PR defines and populates the expanded metrics.
• changed — the per-resource update flag controlling dataVersion, finality reevaluation, and changed-versus-unchanged statistics; this PR extends it to KV mutations but misses dynamic-field removal.
• metadataChanged — the classifier separating structural metadata churn from value/flag-only changes; this PR introduces it for metrics but does not mark field-type transitions.
• traverseStopRules — backend filters that stop traversal at configured resource boundaries; the initial stream applies them, while the new stop-marker follow-up loop deliberately omits them.
• stopMarker — a streaming frame identifying a resource whose traversal was stopped and whose state may require a follow-up fetch; this PR adds multi-round resolution and related counters.
• finalResources — locally immutable resources excluded from additional retrieval; this PR adds statistics around skipped final resources and uses the set while processing follow-up streams.
• KV mutation — addition, replacement, or deletion of resource key/value data; this PR now marks these operations as resource changes so versions and finality are updated.

%%{init: {'theme': 'neutral'}}%%
flowchart TD
  A[Construct loading request] --> B[Initial resourceTree with stop rules]
  B --> C{Frame kind}
  C -->|Resource| D[Collect resource state and metrics]
  C -->|Stop marker already final| E[Skip]
  C -->|Unknown stop marker| F[Queue follow-up seed]
  F --> G[Follow-up resourceTree without stop rules]
  G --> C
  D --> H[updateFromResourceData]
  H --> I{Resource changed?}
  I -->|Yes| J[Advance dataVersion and reevaluate finality]
  I -->|No| K[Count unchanged bytes]

Fix the following 3 code review issues. Work through them one at a time, proposing concise fixes.

---

### Issue 1 of 3
lib/node/pl-tree/src/sync.ts:395-398
**Follow-ups bypass traversal boundaries**

When the initial stream returns a non-locally-final stop marker for a terminal resource with referenced children, the follow-up omits `traverseStopRules` and traverses beyond that resource. This downloads and materializes descendants outside the configured boundary, potentially cascading through a large subtree and defeating the intended load reduction.

### Issue 2 of 3
lib/node/pl-tree/src/state.ts:691-700
**Field removal misses changed state**

When an existing resource loses a dynamic field without another simultaneous mutation, this branch deletes the field but leaves `changed` false. The update is consequently counted as unchanged, `dataVersion` is not advanced, and finality is not re-evaluated until a later mutation.

### Issue 3 of 3
lib/node/pl-tree/src/state.ts:614-643
**Type changes masquerade as stable metadata**

When a Dynamic field transitions to another permitted type, this branch sets `changed` but not `metadataChanged`. The aggregate classifier therefore increments `metadataStableChanged` for genuine structural churn, overstating stable-metadata re-streaming in `MI_LOG_TREE_STAT`.

_{Reviews (1): Last reviewed commit: "MILAB-6303: fetch stop-marker follow-up ..." | Re-trigger Greptile}

Greptile also left 1 inline comment on this PR.

Context used:

• Knowledge Base — pl-tree