this article also has more on which scenarios are the best fit, with a link to a demo repo at the end

Great article Chad!

FWIW, we're working on a faster version of ^%GSIZE (with an API more fitting the current century ;-) ) that uses stochastic sampling similar to the faster table stats gathering introduced in 2021.2? I'll also take the opportunity for a shameless plug of my SQL utilities package, which incorporates much of what's described in this article and will take advantage of that faster global size estimator as soon as it's released.

Please do NOT make any assumptions about system globals such as the ^odd series. User code should stick to documented APIs such as the Export and Clone options described earlier in the thread.