Quoth David Bremner on Nov 29 at 8:17 pm: > Mark Walters <markwalters1009@gmail.com> writes: > > I don't know how freeze/thaw work but does it matter that you don't thaw > > when there is an error? > > My interpretation is that by not thawing before we destroy the message, > we are aborting the transaction, since the freeze/thaw information is > stored in the message structure. It is documented as forbidden to call > thaw _more_ times than freeze, but less is not explicitely mentioned. Yes, this should work. It smells a little hacky and I couldn't say if it was an intended use of the API, but it should work. Perhaps we should document it.