In February 2014 (over three years ago), Vitalik wrote in a blog post that:
Note that there is one weakness in this protocol: you now need to know all transactions ahead of time before processing a block, and adding new transactions requires substantial effort to recalculate intermediate stack trace values, so the process of producing a block will be more inefficient. However, it is likely possible to patch the protocol to get around this, and if it is possible then BP2.1 will have such a fix.
Has such a fix been produced?
