Anatoly Yakovenko told an applauding Consensus Miami 2026 conference earlier this month that Solana would be upgrading to Alpenglow, heralding a live testnet within a week as evidence that it’s “basically due sometime this year, I think next quarter.”
According to developers implementing that upgrade, Alpenglow actually broke while going live.
Anza, the Solana development outfit leading the rollout, left that error out of the mainstream press cycle.
Alpenglow is supposed to be the largest consensus change in Solana’s history, and is supposed to replace proof-of-history which has cryptographically ordered Solana’s network since launch.
Validators approved it in September 2025 with 98% voting in support. Anza, the dev shop, marketed the upgrade as a “100x” improvement to transaction finality, allegedly cutting transaction finalization from 12.8 seconds to about 150 milliseconds.
Yakovenko spoke at Consensus Miami on May 5, praising the potential of Alpenglow. Days later, as promised, it went live on a cluster — and promptly broke.
That day, Anza spokesperson Max Resnick broke the news about the testnet launch to Decrypt, claiming it was “a really exciting milestone” while omitting that it had actually failed and restarted.
On the Solana Foundation’s May 14 weekly validator call, Anza engineer Ashwin Sekar admitted to the real story. Over 40 nodes joined the May 11 cluster running Alpenglow.
Then the migration broke.
‘The next go-around, we were able to successfully perform’
“As usual, the first try did not work,” Sekar admitted on a May 14 call to a niche audience.
He went on to describe the incident, “There was a bug in, you know, the most recent master commit of TowerBFT and proof-of-history.”
TowerBFT is Solana’s proof-of-history-compliant consensus algorithm.
Anza pushed a hotfix and tried relaunching. Sekar added that engineers were hard at work, “We patched it.
The next go-around, we were able to successfully perform the migration.” He also disclosed a second bug, a validator that accidentally banned peer connections, which Anza patched as well.
None of those errors made it into mainstream press cycles.
CHART: Solana survived six years of near-death experiences
Solana could upgrade beyond Proof-of-History by September
Outlets covering the May 11 activation, including TheStreet, Decrypt, CoinMarketCap, and many others reported it as a clean success. Anza’s initial framing of a well-executed migration carried through the news cycle unchallenged.
Coverage of the bug existed mostly within a small audience on YouTube who bothered to watch a recording of a validator meeting.
Yakovenko’s next-quarter claim, which he made in early May, implies a Solana mainnet activation of the Alpenglow upgrade by September 30, 2026.
Solana’s mainnet has halted multiple times in its six-year history, including at least four outages within one 12-month period.
