Follow-up to #2136

Sergio: "I still feel the need to test this via blackbox techniques, as the UTs are mocking a lot of things.
I'm approving this PR now, and will try to drive #2283 to completion so that we can test the rollback command in CI (e2e) with constant changes in validator sets and consensus params."

• refactor(e2e): change *Node to string in keys of some maps #4141
• e2e(main): Tests failing due to validator oscillation #4492
• feat(e2e): introduce oscillations for validators and consensus params #2283
• feat(e2e): add new rollback perturbation #4284