Solana's Alpenglow Switch Completes, 446 of 654 Validators Certify

Solana replaced the consensus system it has run since 2020, and the switch itself took under a second. 446 of 654 participating validators certified the starting block, by CoinDesk's count published early Thursday. The Solana Foundation puts the handover at about 900 milliseconds. Alpenglow now targets finality, the point at which a transaction can no longer be undone, at roughly 150 milliseconds. Every account read here calls that a target. None reports a measured figure from the live network.

Two ways of counting who agreed

The certification number and the activation threshold are not the same measure. Bitcoin.com reported that the migration needed validators holding at least 82% of staked SOL before the network could complete the transition, a stake-weighted test. CoinDesk's 446 of 654 is a headcount of participating validators. No account read here bridges the two, and none explains what the other 208 were doing. The handoff had been pinned to slot 459,653,000 instead of a clock time, which is how a network with drifting block times schedules anything precisely.

Votes left the ledger

Under the old system every validator wrote its votes on chain as ordinary transactions. Votor, the component that went live, has validators signal each other directly and folds many signatures into one certificate, so a block can be finalised after one or two rounds. CoinDesk's dashboard note is the practical consequence: validator votes no longer appear as blockchain transactions. Reported throughput will therefore fall without anything slowing down. Robin Nordnes, chief executive of the Solana infrastructure firm Raiku, told CoinDesk the speed gain matters less to professionals than the headline suggests.

"For desks, faster finality is nice, but it was never the bottleneck. Being close to where the block is built still is," Nordnes said.

Fees, smart contracts and wallets are untouched. Execution was never part of this change, and all four accounts read here agree on that much. The before-figure is where they part: CoinDesk gives about six seconds, Bitcoin.com 12.8 seconds, which the Solana Foundation turns into a cut of roughly 99% and CoinDesk into a factor of about 40. This desk recorded on Wednesday why both baselines are defensible, and the disagreement has not moved since.

Six wipes and one rollback got it here

The path to Wednesday was not clean. CoinDesk reports that the test network was wiped and restarted from scratch six times since early May, and that one migration failed network-wide and had to be rolled back before a later attempt held. Galaxy Digital, one of about 100 test-network validator operators, published its own account of the work. Validators needed Agave 4.3 to make the move, which left the independent Firedancer and Frankendancer clients out of the initial migration; Cryptopolitan reported before the event that their operators would have to run something else for a time. Whether they have switched back is not established in anything published since.

Rotor has no date, and the validator set keeps shrinking

Alpenglow has a second half that did not ship. Rotor, which rewrites how blocks are passed between validators, has no activation date in any account read here. The set it will eventually serve is getting smaller: Bitwise counted 659 active validators at the end of September, down from 696 three months earlier and about 30% below a year before that, with 69% of SOL staked. Concentration moved the same way. The smallest group able to control more than a third of staked SOL fell from 20 validators to 18, and the top ten's share rose from 22.2% to 24.5%. A consensus that finalises in two rounds depends on who is in those rounds.