Lab - Ethereum validators

Staking is an uptime problem, not a crypto problem

Five Ethereum validators, run from 2023 to 2025 on VPS infrastructure. No pool, no staking service: our own keys, our own nodes, our own pager.

A validator earns by being there. Miss your slot and you are penalised; sign twice and you are slashed and ejected. Everything below exists to prevent those two outcomes.

Client diversity, on purpose

Running the same execution and consensus client on every validator is the cheap option and the wrong one. A consensus bug in a dominant client does not just cost you rewards, it can take a share of the network down with it, and correlated failures are penalised hardest.

So the fleet was deliberately mixed across execution and consensus layers rather than standardising on one pair. It costs more to learn, more to monitor and more to upgrade. That is the point.

What the job actually is

• Disk before anything else. An Ethereum node grows without asking. Pruning is scheduled and watched, because a full disk is an offline validator.
• Upgrades on the network clock, not ours. Hard forks are not optional and not negotiable: clients have to be current before the fork, on every node, without a signing gap.
• Never sign twice. Migrating a validator between machines is the one operation that can end it. Slashing protection databases move first, the old machine is proven dead second.

Results

Zero downtime across the run. Five validators, thirty block proposals over eighteen months, no slashing event.