test(sync): preserve bounded Core discovery and priority invariants

[branarakic]

Outcome

Preserve the bounded Core discovery, wake-up, and ordering scenarios from the retired M1 stack in the current RFC64 catalog/coverage architecture.

This is acceptance-first. It must not revive the removed core-public-coverage-scheduler.ts or promote partial discovery without authoritative completion evidence.

Scenarios to preserve

1. When a bounded automatic public scope expands after discovery, the responsible current scheduler wakes promptly instead of waiting for an unrelated peer-staleness window.
2. Scheduler-vetted public VM and SWM work can make bounded progress ahead of a large Agents/Ontology backlog.
3. Private/configured graphs never bypass refreshed Agents authorization/key material merely because public work has priority.
4. A partial bootstrap can expose already durable, authoritative discoveries to later coverage without marking an incomplete ontology/discovery phase successful.
5. Backoff-worthy aggregate results and admission deferral are not themselves proof that authoritative discovery completed.
6. Scope expansion does not create an unbounded rescan, concurrency increase, or reconnect fanout loop.

Related current work:

• bug(daemon): a rejected chain-discovery scan consumes its run slot, silently downgrading the startup seedFull recovery to incremental for ~24h #2323 tracks a rejected chain-discovery scan consuming the startup full-scan slot.
• Reconnect-triggered background sync amplification can starve foreground catch-up #2052 tracks reconnect/background sync amplification.
• Make priority admission fairness and capacity claims explicit #2054 tracks atomic capacity claims and fairness.

Acceptance criteria

• A written owner decision maps each scenario to the current RFC64 catalog/bootstrap/recovery component.
• Focused tests prove wake-up after a newly durable public discovery and no wake-up from rejected/partial non-authoritative discovery.
• Public-priority tests preserve private authorization and system-control-plane prerequisites.
• Partial bootstrap promotion requires an explicit durable discovery-completion witness, not a generic aggregate result.
• Work remains bounded by configured scope/concurrency and respects current admission fairness.
• Any production fix is split into independently reviewable changes after tests demonstrate the current gap.

Provenance

• Extracted from fix(sync): wake Core coverage after scope expansion #2046, fix(sync): prioritize bounded public coverage #2047, and fix(sync): promote discovery after partial bootstrap #2049.
• The red fix(sync): promote discovery after partial bootstrap #2049 implementation path is intentionally excluded: a generic backoff-worthy result cannot authorize discovery promotion.