Multi-node sync hardening: partition and heal

[antonygiomarxdev]

Objective

Stress-test multi-node sync under partition and heal scenarios and close any divergence gaps in checkpoint and idempotency progression.

Roadmap Mapping

• Phase: phase:3-security-governance
• Milestone: v0.3.0

Scope

In scope:

• Partition simulator tool
• Heal regression suite (split-brain, late peer, replay-after-heal)
• Fix any divergence surfaced by the simulator

Out of scope:

• Cross-region replication

Acceptance Criteria

• Partition simulator tool committed
• Regression suite covers split-brain, late peer, replay-after-heal
• No divergence under simulator runs

Test Plan

• Partition harness in CI nightly