Possible use of wrong epoch data

[andresilva]

While running zombienet tests from polkadot the following issue was observed:

WARN tokio-runtime-worker sync: 💔 Verification failed for block <hash> received from peer: <peer_id>, "VRF verification failed: EquationFalse" 

This only happened on one of the validators and it did not recover from the error. Of the top of my head the only explanation I have is that the node is using the wrong epoch data (e.g. randomness) and therefore fails to verify the VRF that other nodes successfully validated.

Javier tried to replicate this by running the same test case and instead got the following errors on two nodes:

WARN tokio-runtime-worker sync: 💔 Verification failed for block 0xb24c05e0aafe58415a1f70574ca8f6d56faa6b51f97848dc95bef765b0bf3ffd received from peer: 12D3KooWKM1HeSv61ryZwAiBTZnqmass5pYM48k9Z7obzhTbnphm, "Invalid author: Expected secondary author: Public(80f832fba260c1c5ee90d9de2fc0189696acaedd5e82538d920bedfba742e725 (13v6rsLY...)), got: Public(c05970a1bf69d1a0b796101c3676eb148cfbca98f119e358dad1097cd3ee4d44 (15MCmG1k...))."    

The error is not the same but could still be caused by using the wrong epoch randomness to calculate secondary slot assignment.

[andresilva]

The easiest thing to do is to try and run the same test with #11228 and #11476 reverted to see if the issue can still be replicated.

I'm also not sure if the issue happened on a node that already had #11476.