We consider a one-dimensional simple random walk among static soft obstacles. The walk has a certain fixed positive probability to be killed each time it meets one of these obstacles. The positions of the obstacles are sampled independently from the walk and according to a renewal process. The distribution of the gaps between consecutive obstacles is assumed to have a power-law decaying tail. First, we prove convergence in law for the properly rescaled logarithm of the quenched survival probability as time goes to infinity. Then, we prove a localization property for the random walk conditioned to survive for a long time, in the spirit of one-island theorems established in related models. We identify two possible localization scenarii, depending on whether the exponent ruling the tail of the gap distribution is smaller or larger than one. This is joint work with Fran?ois Simenhaus.