I have read that a study is very unlikely to replicate if p is juuust under 0.05, but very likely to replicate if just over 0.05. The first is a good sign of p hacking, while the second is a good sign of a real effect with a sample size that wasn't quite big enough.
But I think the underlying reason for the push to publish things - and impactful things are easier to publish, is profit. More hireable grads, more tenured professors publishing papers.