nightly-deploy-hardhat fails with Revive::AccountUnmapped

[EnderOfWorlds007]

What

The nightly-deploy-hardhat cell in .github/workflows/e2e.yml fails consistently when triggered (schedule or workflow_dispatch). The nightly-deploy-multi cell — which uses the same fixture pattern (foundry-style out/) — works.

Symptom

▸ contracts…
▸ build…
✔ build
▸ storage-and-dotns…
  📱 Approve on your phone: Fund contract deploy session key (1/1)
[+0.0s]    Account: mapped
  ✖ contracts: Dry-run failed: {
    "success": false,
    "value": {
      "type": "Module",
      "value": {
        "type": "Revive",
        "value": { "type": "AccountUnmapped" }
      }
    }
  }

The preflight reports SIGNER's account (5DfhGyQdFobKM8Ns… / h160 0x35cdb23f…) as mapped, but the contract dry-run still fails with AccountUnmapped — likely because the contract phase generates a per-deploy session key that hasn't been mapped to an h160.

Why it surfaced now

Nightly cells were never actually exercised in CI before today — Phase 5b/5c/5d ran on PR push only (where the nightly matrix is gated out). The first workflow_dispatch run after Phase 6 SIGINT exposed it. The first schedule cron at 06:00 UTC will reproduce.

Hypothesis

The contract-deploy session key needs to be claim-mapped to an h160 before Revive.instantiate dry-runs. Foundry single-contract deploys don't hit this because the session key gets reused across the foundry tests OR because the out/Counter.sol/Counter.json bytecode has no constructor args that exercise the mapping path. Multi-contract works (same fixture style), so the issue is specific to the hardhat artifact's bytecode/abi shape OR the artifact-discovery path in compileHardhat.

Investigation paths

1. Compare extractFoundryBytecode vs extractHardhatBytecode — do they produce different bytes for the same hex?
2. Check whether compileHardhat returns artifacts in a different order or names.
3. Look at runContractsPhase's session-key funding step — is it gated on something hardhat-specific?
4. Run gh workflow run e2e.yml -f simulate_failure=... style debugging if needed.

Workaround

The cell can be skipped from the matrix until fixed. Currently the auto-issue path will fire on every nightly, which is signal but noisy.

Run that surfaced it

• Run 25278015105: https://github.com/paritytech/playground-cli/actions/runs/25278015105
• Phase 6 SIGINT PR (ci(e2e): Phase 6 — nightly-chaos-sigint cell #92, merged) was unrelated; SIGINT cell itself passed in 4.3s.