 Janson lower-tail bound by independent thinning

For increasing containment indicators, attach independent Bernoulli selectors of parameter $t$. Janson inequality applied to the selected events gives $\mathbb{E}(1-t{)}^X\le \exp (-t\mu +t^2\Delta /2)$. Set $t=1-e^{-s}$ and apply the exponential Markov bound. The estimates $s-s^2/2\le 1-e^{-s}\le s$ give exponent at most $-sa+s^2(\mu +\Delta )/2$. Optimizing at $s=a/(\mu +\Delta )$ proves the displayed lower-tail bound, with the dependency sum ordered and excluding the diagonal.