What I was looking for
Whether a result from one session, or one half-year, holds in the next.
What I found
A caveat that is written down but never tested tends to become permanent.
My order-flow study covered one afternoon of one trading day, and even inside it the 120-second markout drifted from 0.564¢ to 0.253¢ to −0.149¢ by the hour. The strategy rebuilds later showed the same thing at scale: only 33% of setups that worked from January to June still worked from July to October.
The move
I name the regime a result came from, and treat "test another regime" as a task, not a caveat.