Problem with using `kfold` with stansurv objects?

Summary:

There appears to be a problem with using kfold on stan_surv models. Error returned is: Error: passing unknown arguments: subset.

Description:

I'm fitting multi-level survival models with time-varying using rstanarm::stan_surv models and want to use kfold cross validation where groups of individuals are heldout (individuals are repeatedly surveyed overtime). I tried aggregating the log likelihoods as done in the time-varying coefficient example in the Arxiv paper (https://arxiv.org/pdf/2002.09633.pdf), however I get pretty poor Pareto k diagnostic values. So I thought using kfold might be a better alternative, where I group individuals into folds. However, when I do this I get the following error Error: passing unknown arguments: subset.
NOTE I also get this error if I don't stratify/group the data into folds, and if I exclude the random effect for individual.
Any advice welcomed!

Reproducible Steps:

Taking the reproducible example presented in the Arxiv paper:

library(rstanarm)
library(survival)

dat <- survival::pbc
dat <- dat[dat$id <= 312, ]

dat <- tmerge(dat, dat, id = id,death = event(time, as.numeric(status == 2)))

dat <- tmerge(dat, survival::pbcseq, id = id,ascites  = tdc(day, ascites),
              bili = tdc(day, bili),
              albumin = tdc(day, albumin),
              protime = tdc(day, protime),
              alk.phos = tdc(day, alk.phos))

dat <- dat[, c("id", "tstart", "tstop", "death", "bili", "protime")]

mod_tvc <- stan_surv(formula = Surv(tstart, tstop, death) ~ log(bili) + log(protime) + (1|id), data = dat)

inds <- loo::kfold_split_grouped(K= 10, x =mod_tvc$data$id)
kfold(mod_tvc, folds = inds)
Fitting model 1 out of 10
Error: passing unknown arguments: subset.

RStanARM Version:

packageVersion("rstanarm")
[1] ‘2.21.2’

packageVersion("loo")
[1] ‘2.3.1’

R Version:

getRversion()
[1] ‘4.0.1’

Operating System:

OSX 10.14.6

Reproduce the supplied stan_surv example with loo::kfold_split_grouped and kfold(mod_tvc, folds = inds), then investigate how kfold handles stan_surv models and the subset argument. Done means the grouped and ungrouped kfold calls no longer fail with “passing unknown arguments: subset.”