The author states in the comment section, "Spamming will become counter-productive with the new system, and I expect noticeably so if your average latency is above 150ms (as you get twice that amount since the signal needs a round-trip)." This is madness. '150ms latency' refers to a 150ms round-trip time, not 300.
I don't believe his diagrams show the whole picture, either. The nature of cast mechanics are largely theoretical since we can't see the code ourselves. The true test of any theory is whether it allows one to make accurate predictions. The one prediction that I see coming from this theory is that "spamming will result in a larger penalty than timing." However, I have anecdotal and measured data from numerous subjects which counters this assertion flatly.
Fact 1: Casting interval as measured by LagInfo | World of Warcraft Addons | Curse (records the time between client-side cast completion and cast start) is drastically reduced in all observed cases using AHK over manual timing. For consecutive casts of the same spell, an interval of 0ms can be achieved reliably.
Fact 2: Manually parsing combat logs in Excel, the interval between CAST_START messages have shown intervals consistently well below round-trip time. Test results here: http://elitistjerks....0-post1654.html
Anecdote 1: Using AHK with a 4:1 spell rotation I was able to achieve 2k DPS as an elemental shaman w/o 4-set bonus (=http://wowwebstats.com/bvswaybvv5rdo?s=15245,15359).
There is no observable "spam penalty."
In my experience I have never achienve an average 0 ms between casts but mine has dropped from 150ish to about 40ms which is still quite nice. Of course it also saved my poor hand from the frenetic button smashing since that would actually lower my healing output by taking my finger off the key.