I don't have time just yet to write a full reply. But obviously this is a complex area, and I've simplified the presentation for a general audience, so perhaps omitted some of the more nuanced results/discussion from the paper.
This has been discussed alot - and even hotly debated on this forum. I tend to disregard this study for a few reasons.
Read it before you dismiss it?
First and foremost, how many plants were utilized in each sample? It appears to be one per variety.
Three replicates per plant per treatment. So 42 plants total.
If that's true, it wouldn't even get us out of the territory of statistical error. There needs to be a much wider sampling, for this to be taken seriously. Or, if it is a larger sampling, that should be quantified.
Yes, larger sample sizes are ideal as they allow for small effect sizes to be detected. But small samples can still detect statistically significant effects that are large.
The argument can be made that the study might have been underpowered to detect smaller increases in capsaciacin at the 2-day mark, but with an alpha set to 0.05, the type 1 error rate is controlled. (i.e. false positives)
Not only that, there is a clear dose-dependent effect here.
Second point... and of the most importance - there are OTHER studies which suggest that the greatest deviation in capsaicin levels, occurs in those varieties with a much lower capsaicin content. (or lower Scoville rating, to keep it simple) A study at the University of New Mexico found that drought stress was most effective in lower heat level varieties - like any NuMex anuum, jalapeno, serrano, etc. That contradicts this study.
From the discussion section
The previous studies that worked with various types in fruit size and pungency of hot pepper revealed that the small-fruited cultivars are less affected by environment compared with the medium and large-fruited cultivars (Gurung et al., 2012, Gurung et al., 2011, Phimchan et al., 2012, Phimchan et al., 2014).
And from Primchan et al., 2012
However, in our study, the significant increases in capsaicinoids were observed only for the low and medium pungent cultivars but not for the high pungent cultivars. These results might be explained by the fact that a genotype and genotype–environment interaction affected capsaicinoid content (Zewdie and Bosland, 2000), in which the genotype effect was larger than the environmental effect (Gurung et al., 2011b; Zewdie and Bosland, 2000). In addition, Gurung et al. (2012) found high stability of cultivars with high pungency, whereas the lower pungent cultivars were very sensitive to environment. From our results, it is notable that the fruits of the high pungent cultivars are smaller than the other cultivars used, and this might indicate that the capsaicinoids in the small fruit of the high pungent cultivars are less affected by drought stress than those of the big fruit with low and medium pungency.
I might also add the Bosland, head honcho at UNM, is a co-author on these papers.
And I know that the NM study was quite a bit more comprehensive, because it involved money being invested for the direct benefit of commercial growers. (who wouldn't want to make a hotter pepper, and use less water - reduce costs, maintain desirability of product, right?)
This reasoning appeals to common sense but is a bit wishy washy.
You could make a similar argument here, this study is for the benefit of the pharmaceutical industry, who want to extract the optimal amount of capscaicinoids out of their harvests are possible.