That in itself wouldn't be that bad. At least you have a recovery process. It should also be possible to checkpoint the database to a backup periodically and then you only have to replay since the checkpoint.