The effective sample size summarizes the precision lost to reweighting. It is
not a validity diagnostic: a healthy ESS says the weights are well
behaved, not that the reweighted edge is relevant to your estimand. Use
edge_influence() to ask whether the IPD informs the contrast at all.