[codex] split network split playbooks

[parithosh]

Summary

• Rename the slot-level network split playbook to two-way-network-split-reorg-trigger
• Add an epoch-based two-way-network-split-non-finality playbook that preserves the unfinalized-epoch assertion with a 2-epoch floor
• Regenerate the playbook index so both variants are discoverable

Validation

• make generate-playbook-index
• go test ./...
• jq sanity check for the non-finality epoch floor expression

Notes

• The reorg-trigger variant uses splitDurationSlots and only requires fresh finality after healing.
• The non-finality variant uses splitDurationEpochs and enforces non-finality before healing.

[qu0b-reviewer]

🤖 qu0b-reviewer

Summary

This PR splits the two-way-network-split.yaml playbook into two distinct tests: two-way-network-split-non-finality.yaml (epoch-based, asserts finality stops) and two-way-network-split-reorg-trigger.yaml (slot-based, focuses on reorg triggers). Both get registered in _index.yaml. The changes are generally coherent; the substantive fix for minUnfinalizedEpochs with values < 2 is sound. One minor concern remains.

Issues

• 🟡 playbooks/_index.yaml:2,4 — The generated timestamp comment and YAML frontmatter both changed (2026-05-19T21:36:30Z → 2026-05-21T07:03:46Z). These are auto-generated by scripts/generate-playbook-index.sh and will produce spurious diffs every time the index is regenerated. Consider either using a stable/constant value (e.g. generated: stable) or instructing the generator to skip writing this line if its only purpose is documentation.

• 🟡 playbooks/dev/two-way-network-split-reorg-trigger.yaml — Extraction of get_consensus_specs removes SLOTS_PER_EPOCH as a dependency, which is correct for a slot-duration-based playbook. However, if this playbook is ever repurposed to use epoch-based duration in the future (e.g., by adding a splitDurationEpochs configVar), the get_consensus_specs task won't still be there and the jq expression will silently produce no result. This is an internal consistency trap worth documenting in the config section comment.

Suggestions

• playbooks/_index.yaml:266 — The renamed playbook entry (id: two-way-network-split-reorg-trigger) is adjacent in the diff to the new non-finality entry, but in the actual YAML file they remain ordered non-finality then reorg-trigger. Confirm this ordering is intentional; if alphabetical by tag group is preferred, verify the _index ordering aligns with the playbook file ordering.

---

_{Reviewed @ b8ef5961
"If you can't measure it, you can't improve it." — Peter Drucker}