test(e2e): add no-restart back-migration variant + refactor shared scaffolding

[barakeinav1]

Summary

Adds a second back-migration test that does not kill+restart A0 between the forward and back directions, and refactors the shared scaffolding so the two tests share their setup/teardown.

Why a no-restart variant

migration_service__should_handle_back_migration_a_to_b_to_a (existing) kill+restart's A0 to model the disk-persisted-state scenario investigated in #2121. That kill+restart is also the trigger for the upstream nearcore indexer panic surfaced via #3366 — see the diagnostic report for the chain/indexer/src/streamer/mod.rs:207 expect("receipt must be present at this moment") panic. On branches that produce more sandbox chain activity than plain main, A0's post-restart indexer catch-up hits that panic and the test flakes.

Until the upstream nearcore fix lands, this PR adds migration_service__should_handle_back_migration_a_to_b_to_a_no_restart — a stable sibling that exercises the back-migration code path through A0's in-memory state transitions, with no kill+restart and no restart-time catch-up.

What each test covers

Test	Coverage	Triggers nearcore panic?
…_a_to_b_to_a (existing)	back-migration code path + #2121 disk-persisted-state scenario	yes, on branches with high chain activity
…_a_to_b_to_a_no_restart (new)	back-migration code path (in-memory state transitions only)	no

Refactor

The two tests previously had ~85% identical bodies (cluster setup, sanity asserts, forward A→B, post-forward sign+ckd, back B→A, kill B0, final sign+ckd). I factored that into run_back_migration_round_trip(port_seed, between) with a small BetweenDirections enum (None vs KillAndRestartA). Each #[tokio::test] is now a one-line wrapper that picks the port seed and variant.

The kill+restart-specific block (keyshare snapshots, indexer-height snapshot, kill/start/health/indexer-wait, keyshare-preserved assert) lives inside the KillAndRestartA match arm.

Test plan

• cargo check --tests -p e2e-tests clean
• Both tests pass in CI (the new no-restart should be stable; the existing one's flake behavior is unchanged — still tracked under E2E: back-migration test flakes on Connection refused after restart #3366)
• If the no-restart variant goes flaky too, we've learned something important about whether the issue is fully kill-specific

[claude]

Pull request overview

Test-only change in crates/e2e-tests. Extracts the ~85% shared body of migration_service__should_handle_back_migration_a_to_b_to_a into a run_back_migration_round_trip(port_seed, between) helper parameterised by a BetweenDirections enum, then adds a new …_no_restart sibling that skips the kill+restart of A0. The new variant exists so that the back-migration code path keeps a stable green test while the kill+restart variant continues to flake against the upstream nearcore indexer panic from #3366. A new MIGRATION_BACK_NO_RESTART_PORT_SEED = 22 is added so the two tests don't share a sandbox port range.

Changes:

• Adds MIGRATION_BACK_NO_RESTART_PORT_SEED = 22 in tests/common.rs.
• Introduces BetweenDirections { None, KillAndRestartA } and run_back_migration_round_trip(port_seed, between) in tests/migration_service.rs.
• Existing …_a_to_b_to_a now wraps the helper with KillAndRestartA; the kill+restart-specific block (keyshare snapshot, indexer-height snapshot, kill/start/health/indexer-wait, keyshare-preserved assert) lives only inside that match arm.
• New …_a_to_b_to_a_no_restart test wraps the helper with BetweenDirections::None.

Reviewed changes

Per-file summary

File	Description
crates/e2e-tests/tests/common.rs	Adds MIGRATION_BACK_NO_RESTART_PORT_SEED = 22
crates/e2e-tests/tests/migration_service.rs	Extracts run_back_migration_round_trip helper + BetweenDirections enum; existing back-migration test now a 1-line wrapper; adds new no-restart sibling test

Findings

No blocking issues. The refactor is behaviour-preserving for the existing …_a_to_b_to_a test (all kill+restart steps stay in the KillAndRestartA arm and run in the same order). The new test follows the project's <system_under_test>__should_<assertion> naming and reuses the existing helpers for sign/ckd. Port seed 22 is unique and sequential with the surrounding constants.

Non-blocking (nits, follow-ups):

• crates/e2e-tests/tests/migration_service.rs:660 — run_back_migration_round_trip is roughly 130 lines, mostly because the KillAndRestartA arm inlines 6 distinct steps (snapshot, assert, kill, start, wait healthy, wait indexer, assert preserved). If a future variant needs to reuse parts of that block, consider lifting it into a small kill_and_restart_a_preserving_keyshares(&mut cluster, a_idx).await helper. Not worth doing for two callers today.
• crates/e2e-tests/tests/migration_service.rs:631 — BetweenDirections::None is fine but a tad opaque at the call site (BetweenDirections::None reads like "no variant"). A name like KeepRunning would make …_no_restart's wrapper self-documenting. Cosmetic.
• PR description's test plan still has unchecked boxes for "Both tests pass in CI" — worth confirming the new no-restart variant has actually been run end-to-end through cargo make e2e-tests (not just cargo check --tests) before merge, since check-only validation won't catch a port collision or runtime setup issue.

✅ Approved

[Copilot]

Pull request overview

Adds a stable no-restart back-migration E2E variant while keeping the existing kill+restart coverage, and refactors shared A→B→A migration setup into a common helper.

Changes:

• Introduces BetweenDirections and run_back_migration_round_trip(...) to share common test scaffolding.
• Adds migration_service__should_handle_back_migration_a_to_b_to_a_no_restart.
• Adds a dedicated port seed for the new E2E test.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated no comments.

File	Description
crates/e2e-tests/tests/migration_service.rs	Refactors back-migration test flow and adds the no-restart variant.
crates/e2e-tests/tests/common.rs	Adds a unique port seed for the new back-migration no-restart test.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.