What happened in the Ropsten hardfork from Byzantium to Constantinople?

Question

What were the consensus issues during the Constantinople hardfork in the Ropsten testnet?

One noted consensus issue is:

geth and parity have been in noticeable disagreement regarding block 4230605. The discrepancy is gas-used in this transaction - parity charged 791587, geth charged 1078987.

Why did they disagree? Were there other consensus issues?

Answer 1

Problem have been solved over here

Quoting nicksavers here

The thing is that if you did a fast-sync of Geth >1.8.17 after the fork at block 4230605, it would have probably skipped the check on that block and assume it's correct based on proof of work.... then break on 4238641. Similarly, if you start fast-sync now, you'll skip fully verifying that block... etc. (explanation here)

The cause was apparently an implementation detail of EIP-1283. Parity kept track of the refund inside each call, but defined it 'unsigned', which made it impossible to become negative

The solution discussed as of this moment is to use the interpretation of Geth. Parity has already solved the issue on their side with an update (openethereum/openethereum#9746). Miners will start mining on the Geth Constantinople chain on Ropsten to make sure it becomes the canonical chain. The Parity update should be able to follow it. If another conflict arises, it will be dealt with at that time.

Progress should be posted on notes, but it's a little bit behind on the latest information.

Since the error had been solved. But if you are facing this issue please re-sync your node from the start.