This is a long-delayed response. Looking at this, the critical flaw seems to be that it assumes that the variables to be updated – alpha, beta0, and beta1 – are independent.
This seems like an extremely strong assumption to bring to sequential Monte Carlo, and one that in general will not be justified.
In the case of PyMC3 this limitation seems to be baked into the Interpolated class, because it is limited to giving an interpolated distribution of a univariate distribution, but I suppose that one might be able to use a DensityDist or something to model multivariate interpolated distributions.
I’ll be looking into this, but if you have any further suggestions, they would be welcome.