Solana's shorter slots kept skip rates flat but raised vote latency
Serguey Asael Shinder2 min readThe Solana Foundation has published a first look at what happened on chain after the network began shortening its slot time. From block 440,208,000, in epoch 1,019, Solana started stepping down from the 400-millisecond slot it had held for about a year toward a target of 200 ms, through intermediate targets of 350 ms and 250 ms.
The main finding is that the skip rate, the share of scheduled slots that do not end up in the final chain, stayed stable and low through the reductions. The foundation reads that as evidence that validators can keep up. Because slots are shorter while skips are no more frequent, stretches with no new block have also shortened: the dominant component of such consecutive dead time moved from more than 1,800 ms to between 1,200 and 1,400 ms. Some geographic handoffs between leaders, Asia to Oceania and Europe to Oceania, show more skips, though the post notes that the 250 ms sample is small, with only 35 observations for Europe to Oceania.
The cost shows up in vote latency, which rose with each reduction, most for validators in Asia and South America, because votes are transactions that must land on chain. The foundation says average latency stays well below two slots and shows no sign of consensus instability, and that the overall probability of losing vote credits fell, with high-stake validators losing proportionally fewer credits than small ones. Shorter slots also cut the longest continuous stretch a single leader controls transaction ordering: the 99th percentile fell from about 3 seconds to about 2, narrowing the window for extracting value from users.
The post advises caution about a 200 ms target before Alpenglow, the consensus upgrade in which votes become direct messages between validators.
What it means
The data supports going faster so far, with a clear bill attached: latency lands hardest on validators far from the stake-heavy centre. The foundation's own conclusion is that the remaining debate about 200 ms is now more economic and philosophical than technical, which is another way of saying the question is who pays for the speed.