test(e2e): unify slashing timing profiles and merge duplicate suites

[spalladino]

Round-2 e2e consolidation, PR 6 of 9 — scope: multi-node/slashing/**. Every asserted behavior is preserved; no assertions were dropped.

Timing profiles (B3)

The directory now uses exactly two named profiles in slashing/setup.ts, with zero free-floating local slot/epoch timing constants in the test files:

• baseSlashingOpts (existing, eth 8s / aztec 24s) — the equivocation suites.
• SENTINEL_TIMING (new: {anvilSlotsInAnEpoch:4, listenAddress, ethereumSlotDuration:4, aztecSlotDuration:8, aztecEpochDuration:2, sentinelEnabled:true}) — the fast 8s-slot timing.

Every adoption below is a pure re-expression: the effective config values passed to setup are unchanged.

File	Profile	Effective value change?
equivocation_offenses (merged)	baseSlashingOpts	no (both source files already used it)
sentinel_status_slash.parallel	SENTINEL_TIMING	no
validators_sentinel.parallel	SENTINEL_TIMING	no
multiple_validators_sentinel.parallel	SENTINEL_TIMING	no
slash_veto_demo	SENTINEL_TIMING	no
broadcasted_invalid_block_proposal_slash	SENTINEL_TIMING + sentinelEnabled:false	no
broadcasted_invalid_checkpoint_proposal_slash.parallel	SENTINEL_TIMING + sentinelEnabled:false	no
data_withholding_slash	SENTINEL_TIMING + sentinelEnabled:false, aztecSlotDuration:12 override	no
attested_invalid_proposal.parallel	SENTINEL_TIMING + sentinelEnabled:false, aztecSlotDuration:36 + committee:3 override	no

Documented overrides: data_withholding (12s slots for the tolerance window) and attested_invalid_proposal (36s slots so an AVM-heavy 3-block checkpoint fits one slot) keep their bespoke aztecSlotDuration as a documented one-knob override on SENTINEL_TIMING; their eth/epoch come from the profile. Non-sentinel offense files spread SENTINEL_TIMING and set sentinelEnabled:false (= the prior default), reusing only the fast timing.

Kept bespoke: inactivity_setup.ts retains its CI-conditional timing (eth = CI ? 8 : 4, aztec = eth*2) — it is a shared fixture with a deliberate CI slot-doubling that SENTINEL_TIMING's fixed 4s cannot express; the two inactivity test files themselves declare no timing constants.

Merges

• duplicate_attestation.test.ts + duplicate_proposal.test.ts → equivocation_offenses.test.ts: a shared runEquivocationScenario runner plus two describes (the two setups differ in slashDuplicateAttestationPenalty / attestToEquivocatedProposals, so each keeps its own beforeEach).
• inactivity_slash.test.ts and inactivity_slash_with_consecutive_epochs.test.ts stay as separate files. They share the InactivityTest fixture (inactivity_setup.ts), but the fixture is stateful across epochs, so each file still needs its own beforeAll — folding them into one file with two describes would have grown the file without sharing any setup.

Helpers / cleanups

• Centralized findSlashOffense (and the SlashOffense type) in setup.ts, adopted by attested_invalid_proposal and broadcasted_invalid_checkpoint_proposal_slash (the latter's local offense-lookup wrappers now delegate to it, switching validator comparison from string to EthAddress.equals).
• slash_veto_demo: deleted the dead shouldVeto:false branch (it.each([[true]]) → plain it). No running assertion was dropped — the false branch's balance-decrease assertions were never exercised.
• sentinel_status_slash its 1-2: hoisted onto a shared runProposerFaultScenario helper as thin unrolled its (stays .parallel, plain string titles, no it.each).

Verification

yarn build, yarn format end-to-end, yarn lint end-to-end all pass. Ran locally (green): equivocation_offenses (both its ×2), slash_veto_demo, sentinel_status_slash (3), attested_invalid_proposal (2), broadcasted_invalid_checkpoint_proposal_slash (3), broadcasted_invalid_block_proposal_slash, data_withholding_slash, validators_sentinel (4), multiple_validators_sentinel.

inactivity_slash and inactivity_slash_with_consecutive_epochs cannot run in this local env — the InactivityTest fixture derives 0 blocks/checkpoint at the local aztec=8 (default 3s blocks); confirmed the unmodified base files fail identically, and CI's aztec=16 derivation is fine. Relying on CI for those files.

[AztecBot]

Flakey Tests

🤖 says: This CI run detected 1 tests that failed, but were tolerated due to a .test_patterns.yml entry.

\033FLAKED\033 (8;;http://ci.aztec-labs.com/97adeeaef29dbd4e�97adeeaef29dbd4e8;;�):  yarn-project/end-to-end/scripts/run_test.sh simple src/multi-node/slashing/broadcasted_invalid_block_proposal_slash.test.ts (127s) (code: 0) group:e2e-p2p-epoch-flakes