Just wanted to clarify the code regarding regarding:

Make sure state machine doesn't block on channel write if notification handling is slow

The double channel and unbounded slice solution solves this, because if we ever hit the capacity of our inbound notification channel it would block.

For our use cases we would currently never come close to hitting the queue limit given our state machine configurations, and we could easily increase the channel buffer to 1k or more.

But this unbounded solution takes care of it in all cases. I am just wondering about the complexity tradeoff.

@ingar yes... though, keep in mind that single channel is for all deals, so maybe? I dunno. I'm always hesistant to use channel buffers to make a channel into effectively an asynchronous mailbox. but that could be my neurosis.

Oh right, for all deals. That makes way more sense.