Currently when a round needs an echo-broadcast, we run it as a separate round. Theoretically, we could send those messages alongside the messages from the next round. But there are multiple possibilities that will complicate the execution logic:
There may be no next round at all, so we need to run another round anyway just for the echo;
If there's a failure in the next round, the evidence may require the echoed messages. So we need to make sure that we finish the echo round and collect those messages before constructing an evidence;
The list of destinations for the previous and the next round may be different, so some nodes will only get the echo round messages, some will get the new round messages, and some will get both.
Maybe there's something else I haven't thought of. This will make the logic much more complicated for, honestly, a rather modes benefit. Or at least that's how it seems to me now. I'll leave this open for a while in case someone comes up with a way to implement it without a combinatorial explosion.
Currently when a round needs an echo-broadcast, we run it as a separate round. Theoretically, we could send those messages alongside the messages from the next round. But there are multiple possibilities that will complicate the execution logic:
Maybe there's something else I haven't thought of. This will make the logic much more complicated for, honestly, a rather modes benefit. Or at least that's how it seems to me now. I'll leave this open for a while in case someone comes up with a way to implement it without a combinatorial explosion.