Variant of GaussianRandomWalk distribution?

Or maybe this is a silly idea. Maybe the approach is to have a latent GRW for regularly spaced time points, but which you just have observations for irregular time points?