In the thread about Arcane Specs I asked if stopcasting arcane blasts could result into casting the next AB before the debuff of the finished but stopcasted AB is applied and if that could result in a slower casttime.
I set up the following test-scenario:

Dr. Boom, no passive haste-rating, no items with haste-rating procs, blizzard-reported lag about 100ms, Quartz lag something like 100-300ms
modified Quartz to include a debugmessage when it receives "UNIT_SPELLCAST_START" which looks like this:
"casting <Spellname> time <castlength>, start: <startTime>, lag: <lag>"
with castlength=endTime-startTime, lag=startTime-lastt and lastt=endTime(of last cast)

how it should look like (and often does, but see below):

No Stop-Casting, results are okay:


Now it comes, these screenshots are also done while NOT using stopcasting, but sometimes the debuff-information arrives at the client to late.
2 different examples:



I did 20 ramp-ups doing a scorch followed by 4 ABs (no stopcasting), these are the results:
2x 2.5 casts: 5 times
2x 2.1 casts: 4 times
2x 1.8 casts: 3 times
2x 1.4 casts: 8 times (<- this ist the 'correct' behaviour)

Now i tried it with stopcasting...

missing 2 debuffs in one rampup:

or like this



my results:
1. casttimes for Arcane Blast are 2.5, 2.166, 1.832, 1.498 seconds
2. rampup is broken about 50% of the time resulting in losing 0.334 seconds
3. rampup with stopcasting is even more broken resulting in a lot slower casttimes

Did I have some wrong assumptions? Could the tests somehow be improved? How do these results map to normal bossfights (most likely more lag)?

And what about conclusions? Should we wait something like 0.1 seconds longer while rampup-time to make sure that the debuff arrives (and not bashing the key to start the cast as fast as possible)? Can we model these problems for arcane rotations like 3xAB 1xAM 1xScorch?
How much dps are we losing here (assuming no lag the rotation is something like 13 seconds long, so losing 0.3 seconds means 2.5% less dps?

Of course the client is confused. It's in a perpetual state of confusion (thus our use of stopcast).

Did you experiment with stopcasting the 2nd, 3rd & 4th ABs earlier than your lag ought to allow?

What I don't really get is this:
There are two messages sent from the server "your cast has finished" and "you gain Arcane Blast (1)". Missing the debuff without stopcasting (what is actually happening) means starting the cast between these two messages. Blizzard could send this messages at the same time, so that the problem never occurs (at least without using stopcasting).

Yes, the two last images show normal stopcasting of all 4 ABs. Results in a lot of missed debuffs. Note that all casts were successful. Canceling even sooner (meaning that the cast before does not go off) would mean that the debuff will never be applied of course.

It seems like at least one part of the cycle got messed up 65% of the time.

Correct, it's definitely not an isolated event. More than half of my ramp ups without stopcasting contained 1 missed debuff. I know that the sample size is very small at the moment, but I wanted to post what I already have in order to find any flaws and get some feedback etc...
The problem is that it may depend a lot on your lag and the way you are starting the next cast (waiting until you get the visual response that the cast finished vs. mashing on the key even before the cast finished). As I'm quite new to the arcane play style I used the rotations without stopcasting for a few weeks now. But to compensate for that I got into the habit pressing the key a few times very quickly until the next cast goes off.
Ah well, back to Dr. Boom i guess...

Stirius, thanks for putting my suggestion into action. Really weird results there, not sure what they mean in regards to my normal rotations. Makes me very anxious to see Patch 2.2 just so I can see Patch 2.3 on the PTR and get an idea how that latency fix is going to work and whether it will fix the behavior you saw with AB.

As for AB, since the mechanic seems somewhat broken, would it be better to remove /stopcasting from that one spell and leave it on the others? I was thinking about doing that just to better be able to take advantage of MSD procs but not sure how much dps would be lost to not using the macro vs the added dps from properly utilizing the procs with AM. With the info you are showing that dps lost may be more of a wash anyway since missing a debuff during rampup will cost dps too (although I guess it would be nice for dpm).

Were the results a lot more consistant not using the /stopcasting macro then when using it? or were the inconsistancies pretty common either way?

This should probably get cross-posted in some form to the official Bug Report forum, if it's not already there.

alright, one hour at Dr. Boom later.

Test 1: no stopcasting
50 4xAB rampups with 150 checks resulting in 37 lost debuffs (a few times 2 in one rampup) = ~25% missed debuffs

Test 2: with stopcasting (of course no cast lost)
50 4xAB rampups with 150 checks resulting in 72 lost debuffs (a few times 3 in one rampup) = ~48% missed debuffs

conclusion?
1. A 3xAB 1xAM 1xScorch without stopcasting has a 50% chance to lose 0.33 seconds resulting in an dps loss of about 1.5%. (assuming no lag, is this calculation correct?)

2. Using stopcasting I gain about 0.2 sec with each AB with an additional chance of 23% to lose 0.33 seconds. As a result I'm still gaining about 0.125 seconds through stopcasting for one AB. Thats about 0.25 seconds per rotation - not bad.

What I'm thinking would be to do some combination of /stopcasting and not /stopcasting in the same macro so that I would /stopcast when spamming full out AB and not /stopcast when I was ramping it up or using it during a 2AB-2AM rotation or something of that nature.

I may lose a bit of casting time not /stopcasting between the 2 ABs but since a lost debuff is also a substantial casting time loss I think they may wash themselves out. Also since AM's macro has /stopcasting in it I shouldn't lose any additional casting time there.

When spamming AB though not using /stopcasting would really hurt dps so what I am thinking is a macro like

#showtooltip
/stopcasting [modifier:shift]
/cast Arcane Blast

Only thing I'm worried about is seemlessly making the transition between a /stopcasted spell and one that doesn't use it since that can really mess with timing. Back when I was frost I used a macro that had stopcasting in place normally and then removed it when the elemental was in play (I used to like being able to retarget him while still spamming frostbolt without adding any additional bindings) and it really played havoc with my timing. Well guess I can chuck a few mana bars at Dr Boom and see how it works out.

Apparently we encounter here good old issue of procs occuring with delay (estimated to be 0.5 sec) after their triggers. Things simply refuse to occur instantly. So if server starts new AB before 0.5 sec after finishing previous one you get the bug. In this case in 2.3 it's going to happen all the time in case AB spam.