Fault tolerant staking is a new way of looking at staking. It would be a solution to the current staking. As we all know, any formula or functionality has its faults. In the case of staking, there are some that would have to be modified, for which a solution needs to be found. For example, at present, there is some difficulty in running a backup system for a validator in a secure way because it is likely to explode and its validators will be misinterpreted or misjudged because the client's backup could end up online at the same time and sign something twice. Why? Because there is a model in place that sometimes consists of two independent validator systems sharing the same private key, which are meant to operate at different times. This means that your validator key would be operating in both places, and the two messages could be sent saying different things to different parts of the network for a given task. The network interprets this as an attack on your ability to reach consensus, and punishes you severely for it.

The best solution to this so far has been to not run any backups, but clearly we cannot rely on this forever as there are other problems that could be compounded such as the issue of creating a risk for all sizes of parking operators by not having a fault tolerant system.

Obol is a new project that aims to eliminate technical failure points at validator nodes, opening up a new design space that will enable the next evolution in network validation and parking models. At Obol, a primitive infrastructure called Distributed Validation Technology (DVT) is being designed. DVT enables a new type of validator which runs on multiple machines and clients simultaneously, but behaves as a single validator for the network. This allows the validator to remain online even if a subset of the machines fail, which is called active/active fault tolerance.

Obol's main goal is to enable and empower people to share the responsibility of running the network. This way, if a user is part of a cluster of distributed validators, and his machine dies overnight, the other operators in his cluster will have his back.

Main source: https://blog.obol.tech/tackling-the-staking-problem/