The new client is definitely an improvement and stays much closer to the chain tip, but the logs show that the main problem is still the mining/network gate.
The miner repeatedly reaches 2/2 authenticated peers and becomes ready, then shortly afterwards drops to 0/2 and goes back to waiting. This happens over and over again. In some cases the miner is already preparing a block when the quorum disappears, causing the prepared block to be discarded.
At the same time, block propagation is still not truly real-time. The node frequently has to catch up using multi-block compact suffixes, and there are also cases where it detects a recent gap and has to re-request headers because the next block is not immediately available.
This means the miner is not continuously receiving a fresh tip/template and is losing mining time between blocks. With a 15-second block time, these delays are especially significant: by the time the miner becomes ready again, the network may already be several seconds or even multiple blocks ahead.
The reorg logic appears to work when a competing chain is actually detected, so the remaining critical issue seems to be the unstable authenticated-peer quorum and the delayed/fragmented propagation of new blocks. The miner needs to stay connected and ready continuously, and each new block needs to propagate fast enough that miners have a realistic opportunity to mine the next block.
Combined with delayed block propagation and blocks arriving in batches, this means miners often miss entire blocks before they even get a chance to work on them.
There is also an local RPC error:
local node RPC: error sending request for url (http://127.0.0.1:9401/)Even if the current synchronization and peer-quorum issues are fixed, I think a longer block time would still be the sensible choice for mainnet. A longer interval may hide some of the current problems by simply giving the network more time to propagate and process each block, so 15 seconds is a useful stress test rather than necessarily a good production target.
If the client can reliably operate at 15 seconds on the testnet after the current issues are fixed, that would be a strong result. But for the mainnet release, I think 1–2 minutes per block should be the minimum. It gives nodes enough time to propagate and validate blocks, update state, and gives miners a realistic opportunity to receive the latest template and mine each block.
Parano1d v1.0.1 Latest
@ignotusnemo ignotusnemo released this 24 minutes ago
v1.0.1
1d3a01f