On Tuesday, the BNB Smart Chain (BSC) network successfully activated a major software upgrade known as the Pasteur hard fork. According to the development team, the update is aimed at comprehensively strengthening the security of cross-chain bridges, staking mechanisms, and decentralized network governance systems. The key technical feature of the upgrade is an increase in block capacity without changing the established block production time, which remains at 450 milliseconds.
Technically, the Pasteur upgrade combines three key proposals for improving the network architecture (BEP). First, the BEP-682 standard introduces a strict mechanism for rejecting duplicate validator entries during the verification of cross-chain light blocks. Second, BEP-695 tightens the procedures for controlling validator cryptographic key rotation, applying network sanctions (slashing), and conducting governance votes. Third, BEP-675 modifies the algorithm for transmitting block data from specialized builder nodes to validators. Together, these changes prevent a single validator from being counted more than once when approving bridge operations, automatically revoke the credentials of outdated keys, and block voting rights for network addresses that are subject to restrictions.
Particular attention in the upgrade has been paid to optimizing the block formation process during periods of high network load. Previously, BSC's architecture required the builder to execute transactions before sending a proposed block, after which the validator was required to re-execute the same computations before final signing. As the developers noted, this double processing consumed a significant portion of the time within the strict 450-millisecond block production window, which often resulted in blocks being left underfilled. Under the new BEP-675 standard, builders can now send fully pre-executed blocks. Validators first check the proposed block for compliance with consensus rules, sign and propagate it, and only complete the full transaction execution verification after the fact. The previous operating mode, in which validators execute transactions before signing, remains available as an option.
During testing in the internal QANet environment, which simulates the operation of geographically distributed BSC validators, the new mechanism demonstrated significant performance gains. Network throughput increased by approximately 88%, reaching 2,324 transactions per second compared to the original 1,237 transactions per second. The average gas consumed per block rose from 46.35 million to 84.15 million units. The base block interval and the established gas limit of 100 million units remained unchanged. The developers separately emphasized that these figures were obtained under controlled test load conditions and do not necessarily reflect the final measurements on the mainnet.