For that past month or so, our various fixes and optimizations have done a fairly good job of preventing long blocks and keeping token emissions on schedule. For more details see here and here.
However, we've still not managed to restore our 60s block times. Because of this, we're shipping a firmware soon that will contain a new feature and an adjustment of existing behavior that should allow us to close the gap and to stay much closer to optimal token production rates. Firstly, we're adjusting the catch-up time feature to be more aggressive in its adjustments, which should allow us to recover from long block times more quickly. Secondly, we're introducing a consensus feature which records which consensus members completed the consensus protocol round correctly and quickly, and also which other members each member sees during its own round. Once this feature is activated (look for a chain var soon), we should be able to remove downed consensus members much more quickly and reliably and will more often remove those nodes which are consistently lagging behind due to poor internet connections or faulty hardware. As always, see the PRs below for more details.
Additionally this firmware contains various bug fixes, including an important fix for the Proof of Coverage state machine, which can, in certain circumstances, become stalled indefinitely (or at least until the Hotspot restarts).
- Presence Voting: PRs: miner/340, core/392
- Make Catch-up More Aggressive: PR: miner/348
- Fix PoC Stalls: PRs: core/398, libp2p2/264
This was deployed to the beta group around 6PM PST on Wednesday the 18th and will run until the morning of the 19th, when we'll release it to GA.