A Million Random Digits Was a Number-Cruncher’s Bible. Now One Has Exposed Flaws

I didn't read the paywalled article, but if there were expected to be 40.5 four-match sequences, and 48 were observed, that doesn't seem to be a problem.

A Poisson distribution (i.e., rare events limit) should approximately apply here. For a Poisson with a rate of 40.5, the standard deviation would be sqrt(40.5) ~ 6.4, so 48 counts is just over 1 standard deviation high (1.18 to be exact). Not a problem.

Especially if we were measuring k-matches, for k=2, 3, 4, and looking for an exceptional value in any of the above.