Block time: 4 validators cause ~90% of our slow (~10.5s) blocks Over the last 24h, 113 blocks took ~10.5s instead of ~6s. Each time, the round-0 proposer's block didn't reach 2/3 of validators before timeout_propose (3s) ran out, so the network fell back to round 1. No proposals were rejected; they arrived too late. Four nodes account for 101 of the 113: | Node | Failed / turns | Missed commit sigs | |---|---|---| | thor1a7ut4mcte44aq8mun6lhj3rh73j95m08hfdvsr | 51 / 126 (40%) | 11.9% | | thor17s8u4s635kee8g2u34htqxxg6jalvgvjpdpxsu | 36 / 127 (28%) | 5.0% | | thor1sngd0zz6pwdx2e20sml27354vzkrwa4fnjxvnc | 14 / 127 (11%) | 1.1% | | thor1luvyfs2r4cedmmv0fqr2dqdcr4j4fd764hquu3 | 3 / 127 | 0.5% | Example heights where round 0 failed: 28532911, 28533320, 28533318, 28534549 (more on request). Most validators fail 0 of their turns. The last three nodes share an operator and often fail in pairs two heights apart, which points to something shared in that setup rather than network distance. If one of these is yours, please check: - CPU, disk IO and memory headroom on the thornode host, especially around those heights - timeout_commit, timeout_propose and the p2p settings in config.toml are the defaults (timeout_commit = 5s) - Clock sync (NTP/chrony) - Peer count and connectivity (/net_info), plus any rate limiting or firewall on 26656 - thornode logs around those heights for slow block execution or proposal creation Fixing these would take most of our >7s blocks away. Happy to share the full data or re-run the analysis after changes. 🙏(Historical earnings: For 2026Q2 ended 2026-06-27, Intel reported basic and diluted EPS of -2.89, net loss of USD 14.761 billion, and revenue of USD 29.705 billion. Consensus expectations: For 2026Q4 (event date 2026-10-29), consensus EPS is 0.3953 and consensus revenue is USD 16.473 billion.)