The two probes now differ on purpose, and the seal is enforced where months are read¶

scripts/separability_over_time.py. The review noted that this probe reads every built month, dev included, where separability.py stops at train, and left it. The difference is right and was undocumented: this one asks whether the corpus moves over time, and a drift reading that stops two months early cannot see the end of its own series. It reads corpus text and no held-out predictions, so the dev window is described rather than spent. The docstring now says that instead of implying both probes follow one rule.

What was missing is a guard. The months came from whatever datasets/gerrit/<org>/examples holds, so a test month fetched after acceptance would walk into a descriptive analysis by being on disk. by_month now refuses any month at or past the sealed window and says so, and both readings record sealed_from. Rerun on the committed drift case, the numbers are unchanged: OpenStack 0.552, Qt 0.510, the same early and late slices.