Attack Speed Reference

As little as 600ms, yes. There was a lot of variation while I was mashing keys.

It could be that it rounds it, although I don't think that's it. If it rounded, it would be periodic while I sat there attacking and I was unable to find any sort of consistency or pattern to it.

As for the difference between server-based ticks and client-sent ticks, that's dependent on time of execution.

For all I know, one packet may have needed to be retransmitted a couple times (yes, even to 127.0.0.1) or my processor may have taken longer on one function or data setting along the way because it was busy with something else at the moment. It's not surprising that there's a variation in server-sided tick count.

With the client-sent tick count, you're guaranteed to get uniform numbers for the same thing because it only sends the tick count with packets and it will only send packets as fast as it will allow you to. Theoretically, this should be subject to the same processing pitfalls that the server is. I'm not sure what techniques are employed to remove the variation - although rounding would probably...