Currently, the test network has stopped releasing blocks. I’ll tell you what happened.

When a new light node registers on the network, all nodes must calculate the checksum of the node registry in the same way. It turned out that the node that had itself issued a block with such a registration calculated it differently from the others. It stopped matching the network in terms of the same number and could no longer confirm blocks.

Two light nodes registered. The first one disabled one node, the second one disabled the second one. After that, four out of the five required nodes remained, the confirmation stopped, and the network automatically suspended production. This is a standard safety feature: it’s better to stop than to diverge.

The data is intact. The entire history is in place, and the records are the same for all six nodes. Only one calculated value diverges, and it can be restored by recalculating it.

What I do: I restore agreement on the two affected nodes, fix the cause in the code so that this doesn’t happen again during registration, and update all the nodes.