That is a very good question. I have been wondering that myself.
When 1 of the 4 vaults goes offline, the data is copied to a new vault. And when the old vault comes back online with data previously planted, we have a de-duplication event. This only happens after 10 minutes, called the “back off period.” I’m not 100% sure this is the way it works, so don’t quote me on this. It’s just my best guess understanding.
Vault 1-4 has the 4 copies
Vault 1 = 100Gb
Vault 2 = 100Gb
Vault 3 = 100Gb
Vault 4 = 100Gb
Vault 1 = 100Gb
Vault 2 = 0Gb (Goes offline for more than 10 minutes)
Vault 3 = 100Gb
Vault 4 = 100Gb
Vault 5 = 100Gb (Data recopied to this vault)
Vault 1 = 100Gb
Vault 2 = 100Gb (Comes back online)
Vault 3 = 100Gb
Vault 4 = 100Gb
Vault 5 = 100Gb
This triggers a de-duplication event which the Network will “remove” data from a vault.
I am guessing the vault with the lowest rank will have its data removed. The reasoning is because low ranks tend to go offline more frequently, and thus temporary data should be stored with them. So if we apply the ranks in this example. Then the following should happen.
Vault 1 = 100Gb = Rank 3
Vault 2 = 100Gb = Rank 2
Vault 3 = 100Gb = Rank 2
Vault 4 = 100Gb = Rank 1 (Candidate for De-Duplication)
Vault 5 = 100Gb = Rank 1 (Candidate for De-Duplication)
Now the Network has to determine between Vault #4 and #5 to remove their data. If the Network is also measuring their online time, then it can make that decision. If Vault#5 has been online longer than Vault#4, then Vault#4’s data will be removed.
If a Vaults data is removed via de-duplication, it can be filled up with new data, so the loss should not be too detrimental to the farmer. The important thing is to try and stay online consistently, and for as long as possible. The 10min back off period should give you enough to time to restart your computer before the data is copied to another vault.
If anyone thinks this is the wrong explanation, please feel free to add, adjust, or correct me.