> This suggests either that the negative results were truncated out, negative results have been hacked to positive ones, or there is some confounding factor at play.

Here's a potential confounding factor: maybe the cluster of very weak positive effects includes some studies that initially showed very weak negative effects but that had their methodology tweaked to avoid showing an "impossible" result. (I don't think anyone argues that atmospheric lead _decreases_ crime rates.)

If so, that would mean that there are some "missing" negative studies, but that there are also "too many" zero/very low result studies. If corrected for that, I suspect that the meta-analysis would show a positive effect size, though lower than 0.33.