Year Two alone accounted for 144.76 of the subThreads and 107,490 of the comments (= 79%).
Here's the overview:
Fig. 1: [click to enlarge] Comment count for the first two years of Pharyngula's Endless Thread. The colorful vertical lines mark each of the first 50 or so subThreads, but I gave up on that shit. Red threadmarkers with adherent vestigial green threads mark anastomoses. The pink threadmarkers are the 1- and 2-year marks. Somewhere in there in light blue are the parallel subsubThreads CXXVb and CLXIIb.
Up, up up. It's a long way from last year for damn sure.
Commenting rates over Year Two's subThreads varied from 150 to 813 comments /d (Fig. 2)
Fig 2: [click to enlarge] Commenting rates for each subThread, plotted at end dates. Pink markers delineate years 1 and 2. Black line is a 3-point moving average to de-emphasize individual data. Red curve is a best-fit third-order polynomial.
Pretty variable, but possibly stabilizing at a long-term mean of about 370 comments/d.
SubThreads of Year Two lasted an average of 2.57 d (Fig. 3).
Fig. 3: [click to enlarge] Duration of each subThread. Note logarithmic Y-axis. Conventions as in Fig. 2.
Interestingly, the very first subThread of Year Three was the first ever to come in at less than 1 d.
Finally, the relationship between duration and commenting rate continued to be well described by a power curve (Fig.4).
Fig. 4: [click to enlarge] Log-log plot of subThread duration vs. commenting rate.
I still don't know what, if anything, this means. However, it seems that less than 5% of the total variance in duration is attributable to factors other than commenting rate (e.g., PZ changing the portcullis-target, PZ getting drunk, PZ's general poopyheadedness, etc.).