I would like to see the ability to fit Cox models with time-dependent covariates, as these are very common in retrospective, medical research. While the creation of (start,stop) counting process data sets might be challenging in JMP's point and click environment, perhaps an approach of submitting "wide" data sets to JMP's Cox platform with variables storing the time change-points and the corresponding new values might work (as in Paul Allison's excellent book "Survival Analysis Using SAS").