test(storage): probe reachability of the comparisons state-sync path

[chefsale]

What

Adds three WARN-level, greppable probes (target: \"sync_probe\") at the only sites that can exercise the legacy state-based StorageDelta::Comparisons / compare_trees sync path. No behavior change — probes only log.

• delta.rs commit_root — the sole producer of a Comparisons wire artifact.
• root.rs apply_comparisons — fires whenever any Comparisons delta (including the zero-byte no-op that deserializes to Comparisons(vec![])) reaches Root::sync.
• root.rs apply_actions Action::Compare arm — fires if a Compare action is ever received from peer traffic (the trigger that becomes a Comparisons emission).

Why

We want to delete the compare_trees / Comparisons subsystem (it's deletion-unaware and superseded by HashComparison/LevelWise/DAG-CausalActions). Static analysis strongly suggests it's already unreachable in production:

• Action::Compare is pushed only inside apply_action, into context.actions (not comparisons), and only during __calimero_sync_next — a state-op whose artifact isn't broadcast.
• Authored/stored DAG deltas are always CausalActions built from local writes (Add/Update/DeleteRef), never Compare, never empty.
• So peers' __calimero_sync_next should always receive CausalActions → apply_actions, never the Compare arm, never apply_comparisons.

The one residual uncertainty is whether the node-side state-delta handler ever re-authors/propagates a DAG delta from a sync-apply outcome (which would carry Compare actions to peers). Rather than guess, these probes let a real multi-node run answer it with data.

How to read the result

Grep node logs (guest probes surface via the env::log bridge) for:

sync_probe: apply_comparisons invoked
sync_probe: received Action::Compare in delta
sync_probe: commit_root emitted StorageDelta::Comparisons artifact

Run something that forces Merkle divergence (concurrent merges + deletes across nodes). If zero sync_probe lines appear, the path is confirmed dead and the follow-up PR removes the whole subsystem (and the now-false convergence comment at interface.rs:2043-2049).

Follow-ups (not in this PR)

• Full removal of compare_trees / compare_affective / generate_comparison_data / Comparison / StorageDelta::Comparisons / apply_comparisons, pending a clean run.
• Related: open-invitation multi-claim gap tracked in Open group invitation is multi-claimable by any syncing identity until expiry (no joiner-binding / no single-use nonce) #3207.

🤖 Generated with Claude Code

[meroreviewer]

🤖 MeroReviewer

Reviewed by 1 agents | Quality score: 85% | Review time: 157.6s

---

✅ No Issues Found

All agents reviewed the code and found no issues. LGTM! 🎉

---

_{🤖 Generated by MeroReviewer | Review ID: review-d23891c3}

[meroreviewer]

Documentation Review

The following documentation may need updates based on the changes in this PR:

• 🟡 architecture/migrations.html: Files matching crates/storage/** were changed but architecture/migrations.html was not updated (per source_to_docs_mapping).
• 🟡 architecture/storage-schema.html: Files matching crates/storage/** were changed but architecture/storage-schema.html was not updated (per source_to_docs_mapping).

[meroreviewer]

🤖 AI Code Reviewer

Reviewed by 1 agents | Quality score: 31% | Review time: 273.6s

🟡 1 warnings. See inline comments.

---

_{🤖 Generated by AI Code Reviewer | Review ID: review-1e24cecc}

[meroreviewer]

🟡 apply_comparisons probe fires on the benign zero-byte no-op artifact, undermining the reachability conclusion

The probe is placed unconditionally at the top of apply_comparisons, before the comparisons.is_empty() check. Per StorageDelta's custom BorshDeserialize (delta.rs), a zero-byte wire artifact (the routine no-op that commit_root emits whenever there are neither actions nor comparisons) always deserializes to Comparisons(vec![]), not to Actions(vec![]). That means any ordinary, harmless empty commit that happens to reach Root::sync will trip this WARN with comparison_count=0, even though it has nothing to do with the legacy compare_trees subsystem being investigated. By contrast, the sibling probe in delta.rs::commit_root is correctly scoped to the (&[], _) (comparisons genuinely non-empty) match arm and will not fire for that same no-op case. This asymmetry means a non-zero sync_probe: apply_comparisons invoked log line does not actually confirm that a peer ever sent a genuine Comparisons payload — it could just as easily be triggered by the unrelated, expected empty-artifact fallback. Since the PR's explicit decision rule is "if zero sync_probe lines appear, the path is confirmed dead," a spurious non-zero count here could incorrectly block the planned removal of the subsystem.

Suggested fix:

Split the probe so the zero-byte/no-op case is distinguishable from a genuine received Comparisons payload, e.g. only warn when `!comparisons.is_empty()`, or add a separate lower-signal log line for the empty case: `if comparisons.is_empty() { tracing::debug!(target: "sync_probe", "apply_comparisons invoked with empty comparisons (likely benign zero-byte no-op)"); } else { tracing::warn!(target: "sync_probe", comparison_count = comparisons.len(), "apply_comparisons invoked with non-empty comparisons"); }`

[meroreviewer]

🤖 AI Code Reviewer

Reviewed by 1 agents | Quality score: 85% | Review time: 373.6s

---

✅ No Issues Found

All agents reviewed the code and found no issues. LGTM! 🎉

---

_{🤖 Generated by AI Code Reviewer | Review ID: review-eab9e5bb}

[chefsale]

Superseded by #3215 (the actual removal). This probe branch's purpose — hard-abort instrumentation to confirm, via the multi-node divergence e2e, that the state-based Comparisons sync path is never reached — is complete: no abort fired, so the path is dead and #3215 removes it. Not for merge.