There are (at least) two known problems at the moment:
There are other transactions in a real CCF, and so the txids aren't nicely contiguous as expected by the spec. @heidihoward had a solution to this, adding stub transactions and leveraging non-determinism, but I can't seem to find it right now.
The trace does not emit TruncateLedgerAction, and so traces with a rollback end up with an incorrect CommitSeqNum at the first Invalid response
I am sure there are more, but I am not planning to find out before Monday.
Early draft follow up to #6116
There are (at least) two known problems at the moment:
I am sure there are more, but I am not planning to find out before Monday.