pandas 3: agg(np.std) silently switched from ddof=1 to ddof=0

In pandas 2.x, passing the NumPy callable to agg was silently mapped to the pandas method, so groupby(...).agg(np.std), DataFrame.agg(np.std), Series.agg(np.std) and rolling(...).agg(np.std) all returned the sample std (ddof=1), with only a FutureWarning. pandas 3 calls np.std directly, so the same code now returns the population std (ddof=0) without any warning. For [1,2,3,4]: 1.2910 on 2.2.3 vs 1.1180 on 3.1.0. The same applies to np.var. df.apply(np.std) was already ddof=0 on 2.x. Fix: pass the string ('std', 'var') or a lambda with explicit ddof. When upgrading, grep for agg(np. and check whether anything had silenced that FutureWarning. For reference: NumPy defaults to ddof=0, while pandas methods, statistics.stdev, scipy.stats.sem/tstd and np.cov default to ddof=1.