fix(tests): avoid races and leaked state in tests

[mkoura]

• test_tx_chaining: lock the whole cluster instance. The test checks how many blocks the chained Txs take, so Txs of other tests must not take the block space, and the chained Txs fill the mempool for other tests.
• Plutus script stake address tests: respin the cluster instance when a test leaves the script stake address registered, otherwise all the following tests using the script are skipped.
• KES negative tests: pass a copy of the pool's cold counter file to issue-op-cert, as the command would increment the shared counter if it succeeded.
• test_rotation_pair_keeps_voting: check the epoch length in the fixture, before the respin is requested.
• test_permanent_fork: request the respin before splitting the cluster, so it is requested even when the split fails half-way.
• set_network_id_env: restore the original CARDANO_NODE_NETWORK_ID value instead of deleting it.

[Copilot]

Copilot review overview

🟡 Changes recommended

The KES tests still read POOL2’s shared files without declaring that pool as in use, leaving a concurrency race.

Review effort: Balanced
Findings: 1

Open (1)

• Declare POOL2 use before copying the shared counter · New

What changed in this PR

Improves E2E test isolation by preventing shared-cluster races and restoring modified state.

Changes:

• Adds cluster/resource isolation for transaction and Plutus tests.
• Protects shared KES counters and restores environment state.
• Requests cluster respins earlier and performs epoch checks before mutation.

File	Description
tests_plutus/​test_delegation.py	Respins when script stake registration leaks.
tests_dijkstra/​test_bls_rotation.py	Checks epoch duration before requesting respin.
test_rollback.py	Requests respin before cluster splitting.
test_kes.py	Copies cold counters before negative commands.
test_env_network_id.py	Restores environment and CLI arguments via monkeypatch.
test_chain_transactions.py	Locks the cluster and marks the test as heavy.

---

💡 Add a code-review agent skill or configure MCP servers for context-aware, tailored reviews. Learn more in the docs.