The data for overall survival endpoint has not come close to maturing as a large percentage of the treated group are still alive but censored (as dead) for this second analysis. A quick look at the survival KM shows a very large cohort of survivors in the middle of the treated survival curve. It won't take very much time for these survivors to influence the HR and the p values as the trial matures. The slopes of the curves differ by more than 20% but the censoring of the survivors really influences the data in the middle of the curve.
Ok - but the survival data doesn't actually look all that good. May hit stat sig at survival, may not. Very borderline and not a particularly good HR - although almost certainly active in the disease.
I agree with Dew that the most interesting trial(s) will be those combining it with immune debraking treatments. Hopefully they allow better durability - thus translating the response endpoints into survival.