Currently the code assumes that sample sizes are fixed within a window. But many interesting time series have varying sample size (for example, the YJIT benchmarks). Montgomery discusses variable sample sizes several times, though not for all control charts.