We were not considering the records that were returned from the network if the quorum was not fulfilled. But these should be considered during validation.
What the PR buys us is that:
Quickly consolidate double spends attempts and store them in nodes. Clients were already safe, but nodes can possibly be out of sync for a while and would wait for replication to get them to sync. So it just reduces 1 step for the nodes
Also run spend simulation as its own CI run. The local network was getting overwhelemed.
What the PR buys us is that: Quickly consolidate double spends attempts and store them in nodes. Clients were already safe, but nodes can possibly be out of sync for a while and would wait for replication to get them to sync. So it just reduces 1 step for the nodes