Pierce Lopez

I pushed up another commit, a hacky idea to handle the case of `memQueueSize=0` (or `memQueueSize=1` which it seems like some people might pick with the same idea/motivation), by waiting...

OK, I came up with another strategy, which I think is more efficient, fixes some long-standing potential for undesired behavior with very small `MemQueueSize`, and is pretty nifty. But it...

Went back to the simpler idea, with an additional "started" flag.

With `mem-queue-size = 0` (and this change), the messages should still hit disk, but pass more directly to the channels diskqueues if possible (and if not possible still go directly...

... so yes, agreed with what you wrote, but I think this handles (2) just fine, and only somewhat breaks (1) (mainly the feedback).

> Why don't we just revert #1159 Yeah, I'm amenable to this idea. I was trying to please the no-mem-queue use-case here but agree it might not be worth all...

Updated to only change behavior (compared to latest nsq-1.2.1 release) for deferred messages or paused topics. (But it's still kinda a lot of lines.)

That's correct. `#ephemeral` means there is no disk-queue, only mem-queue. `mem-queue-size=0` means there is no mem-queue. So, there is no queue :)

You may see disk queue files because the _topic_ is not ephemeral, but the _channel_ being ephemeral would still have this problem with mem-queue-size=0. The topic and each channel has...

eh, well, the issue was _introduced_ in nsq v1.2.1 ... will it be "fixed" in v1.2.2 or v1.3.0? I don't know, but I have an unapproved pull-request which attempts to...