BSC gives node operators until 02:30 UTC on August 25 to be running v1.7.7

BNB Chain has set 02:30 UTC on August 25, 2026 as the activation time for the Pasteur hard fork on BNB Smart Chain mainnet, and told every node operator to be running client v1.7.7 before then or risk falling off the network ^[1][2]. The date and the client requirement were published in a blog post on August 14, which leaves operators roughly 11 days of notice ^[3][4].

Pasteur sits under meta proposal BEP-673 and pulls three other proposals into a single coordinated activation: BEP-682, BEP-695, and BEP-675 ^[4]. The first two are security work on the bridge and on the staking and governance contracts; BEP-675 is a throughput change that, by BNB Chain's own account, does not technically require a hard fork to function ^[5]. Bundling it anyway means the performance feature and the security fixes now share one deadline.

The bridge fix is the one worth reading closely. When assets move onto BSC, the chain does not take the origin chain's word for it; it runs a precompile at address 0x67 that counts validator signatures against a known validator set, and the transfer clears only once a supermajority has signed ^[6]. There was no check that each validator appeared only once, so a crafted validator set could list the same signer repeatedly, count that signer's voting power on each appearance, and clear the threshold with fewer genuine signatures than the rule demands ^[7]. BEP-682 strips duplicates before the count ^[8].

BEP-695 addresses a related gap in assumptions. Rotating a consensus key was supposed to retire its authority, and did not: the old key kept validator-admin privileges, rotation could be used to dodge a pending eviction, and blacklisted addresses could still cast signature-based governance votes. All three are now closed ^[9].

The throughput change is a scheduling argument, not a capacity one. Today a BSC block is executed twice, once by the builder to confirm validity and again by the validator before it signs ^[10]. That second pass consumes part of the 450 millisecond window, and whatever the validator spends there is time builders lose to pack the next block ^[11]. BEP-675 lets a builder hand over an already executed block; the validator checks it against consensus rules, signs, broadcasts, and completes full verification afterwards ^[12].

On testnet, validator time in the execution layer fell from 125 milliseconds to 15 milliseconds, an 88% cut that returns about 24% of the block interval to block building ^[13][2][1]. Throughput rose from 1,237 to 2,324 TPS, an increase of about 88%, with the 100 million gas limit and the 450 millisecond interval left untouched ^[14][d5b]. Average block gas used went from 46.35 million to 84.15 million, meaning utilisation moved from roughly 46% of the limit to roughly 84% ^[15][3]. In other words, the binding constraint was the clock, not the gas ceiling. Finality lag held steady ^[15].

BNB Chain flagged that these are controlled-workload numbers that still have to hold under mainnet traffic, particularly when several builders compete for the same slot ^[16]. That caveat matters more than the headline figure. For context, the Fermi fork shipped in January 2026 and cut block times to 0.45 seconds, and the chain has spent the months since holding that pace rather than pushing it lower ^[17].

Watch three things after the 25th: whether real mainnet block gas used lands anywhere near the 84.15 million testnet figure ^[15], whether finality lag stays flat once multiple builders contend for slots ^[16], and whether validators actually complete post-signature verification without incident, since signing now precedes full checking ^[12].