@Arcon:

I have sort of been following it. Mainly only noted that there's different treasure pools for different sets of golems in sky.. >.>

The problem is that all evidence for the multi-roll theory seems circumstantial. It works in some cases, but not in others.

Here's some TH data I have from a few years ago:

Code:
Savanna Dhalmel (Killed 1002 times with TH 0)
       11 dhalmel hide                 [Max #: 2]  [Items/Kill:  0.011]  [Drop Rate:   1.00 %]  [% of Drops:   4.40 %]
       22 giant femur                  [Max #: 1]  [Items/Kill:  0.022]  [Drop Rate:   2.20 %]  [% of Drops:   8.80 %]
      100 slice of dhalmel meat        [Max #: 2]  [Items/Kill:  0.100]  [Drop Rate:   8.98 %]  [% of Drops:  40.00 %]
        3 sprig of papaka grass        [Max #: 1]  [Items/Kill:  0.003]  [Drop Rate:   0.30 %]  [% of Drops:   1.20 %]
      114 tuft of dhalmel hair         [Max #: 1]  [Items/Kill:  0.114]  [Drop Rate:  11.38 %]  [% of Drops:  45.60 %]

Savanna Dhalmel (Killed 1000 times with TH 1)
       19 dhalmel hide                 [Max #: 3]  [Items/Kill:  0.019]  [Drop Rate:   1.70 %]  [% of Drops:   3.39 %]
       46 giant femur                  [Max #: 1]  [Items/Kill:  0.046]  [Drop Rate:   4.60 %]  [% of Drops:   8.20 %]
      350 slice of dhalmel meat        [Max #: 2]  [Items/Kill:  0.350]  [Drop Rate:  27.00 %]  [% of Drops:  62.39 %]
        6 sprig of papaka grass        [Max #: 1]  [Items/Kill:  0.006]  [Drop Rate:   0.60 %]  [% of Drops:   1.07 %]
      140 tuft of dhalmel hair         [Max #: 1]  [Items/Kill:  0.140]  [Drop Rate:  14.00 %]  [% of Drops:  24.96 %]

Savanna Dhalmel (Killed 200 times with TH 2)
        7 dhalmel hide                 [Max #: 2]  [Items/Kill:  0.035]  [Drop Rate:   2.50 %]  [% of Drops:   4.52 %]
       13 giant femur                  [Max #: 1]  [Items/Kill:  0.065]  [Drop Rate:   6.50 %]  [% of Drops:   8.39 %]
       96 slice of dhalmel meat        [Max #: 2]  [Items/Kill:  0.480]  [Drop Rate:  37.50 %]  [% of Drops:  61.94 %]
        1 sprig of papaka grass        [Max #: 1]  [Items/Kill:  0.005]  [Drop Rate:   0.50 %]  [% of Drops:   0.65 %]
       38 tuft of dhalmel hair         [Max #: 1]  [Items/Kill:  0.190]  [Drop Rate:  19.00 %]  [% of Drops:  24.52 %]
(I apparently got distracted from the TH2 and higher testing; should go back and finish that)

From the thread, there was mention of another test on dhalmel drop rates:
Quote Originally Posted by Mojo
It only works well if you completely disregard cases where it doesn't work. Dhalmel Meat, which can double drop, dropped at 100/1000 for TH0 and 358/700 for TH2. ... Only 100 Dhalmel Meat were obtained during that set, so the prediction deviates heavily from what was observed.
Drop groupings (not shown) allow easy verification of drop distributions. For mine:

TH0 gave 10 2x drops; the remaining 80 meats would be singles, with 50 drops in each slot (5% drop rate). However that introduces an extra margin of error, so it would be better to consider it as 100 drops out of 2004 'attempts' (kills*slots), with the assumption that drops for each slot are independant. This gives a CI interval of 4.12%-6.04%.

TH1 gave 80 2x drops, leaving 190 single drops, and 175 drops in each slot (17.5%). Getting CI on 350 drops out of 2000 (kills*slots) we get a CI range of 15.90%-19.23%.

TH2 gave 21 2x drops, leaving 54 single drops, and 48 drops in each slot (24%). CI interval is 20.07%-28.43%.


If we assume TH1 was one reroll, the range of the base drop rate is 8.29%-10.13%, significantly out of the range of the TH0 data.

If we assume TH2 was two rerolls, the range of the base drop rate is 7.20%-10.55%, on par with the result from the TH1 data, but still out of range of the TH0 value. They both converge on a 9% base drop rate.

Which, coincidentally, is the "drop rate" of the TH0 data - the number of kills for which at least one of that item dropped. Possibly an indication that the drop rate for the second slot item is gimped under TH0? EG: 9% base drop rate for slot 1, but only ~1% for slot 2; slot 2 drop rate raised to match slot 1 if TH1 or higher is present. A thought to file away for later.


Dhalmel hair:
9.55%-13.5% @TH0
11.98%-16.29% @TH1 >> 6.18%-8.51%
14.14%-25.03% @TH2 >> 4.95%-9.16%

In no case does the back-calculated drop rate based on multiple rolls fall within the CI of the baseline. There's no chance of multiple drops interfering in this set, either.