Allow re-delegation from inactive validators

[gjermundgaraba]

Summary

Allow a delegator to re-delegate outside of the re-delegation lock period if the validator in question is outside of the active set.

Problem Definition

Currently, there is a disincentive to re-delegate to validators that are very close to the bottom of the validator set because the delegator risks having no rewards for an extended period of time. Instead, the incentive for a low-risk re-delegation is to higher-ranked validators, increasing centralization.

With this feature, we would remove this disincentive and show that decentralization is highly valued.

This has been discussed a little here as well: https://twitter.com/GjermundGaraba/status/1571936468590796806

Proposal

How about:

Add a check when realizing the user is trying to re-delegate away from an already "locked" delegation and if the validator is outside of the active set allow it anyway.

[gjermundgaraba]

If/when this gets accepted, I would be more than happy to work on this issue.

[tac0turtle]

is there any reason this isn't the case @alexanderbez?

[facundomedica]

Would this complicate slashing unbonded validators?

[alexanderbez]

What's a "locked" delegation? I don't see any restrictions or checks in BeginRedelegate that explicitly prohibits redelegation from an inactive validator (that would now be unbonding or unbonded).

[facundomedica]

I think OP is referring to this:

A redelegation object is created every time a redelegation occurs. To prevent "redelegation hopping" redelegations may not occur under the situation that:

• the (re)delegator already has another immature redelegation in progress with a destination to a validator (let's call it Validator X)
• and, the (re)delegator is attempting to create a new redelegation where the source validator for this new redelegation is Validator X.

[alexanderbez]

Ahh transitive redelegation. Yes, there is a reason why we prevented this and we already have an (old) issue for it: #3288

I would close this issue and keep that one I mentioned. I don't have the capacity to work on it but it's definitely a great feature to add for x/staking. Would love contribution on it 🙏

[gjermundgaraba]

@alexanderbez That is actually a slightly separate issue.
#3288 is only concerned about the entire delegation being locked (no matter the number of tokens "redelegate locked"). This issue is about ignoring that lock if the validator is inactive.

[alexanderbez]

Can you explain what you mean by "lock". I'm not understanding the proposal exactly.

[gjermundgaraba]

By "lock" I mean not being able to redelegate.

The proposal is about enabling redelegation from a validator that has fallen outside of the active set, even if the delegator is still in the "lock" period.

The goal is to lower the risk for people to redelegate from high-ranking validators to low-ranking validators (in case the validator they delegate to falls out of the active set).

[gjermundgaraba]

@alexanderbez @marbar3778 pinging on this one as I would like to know if I can do this or not :)

[alexanderbez]

@bjaanes yes, we would like to see the approach! I'm skeptical that it'll be that easy as there are various corner cases to consider, e.g. cliff validators dipping in-n-out of the active set. But I would love to see a solution 🙏