A parallel scope keeps its own fork activation

[noha]

FBBaseExecutor>>activationOf: answers the most recent activation of a node. A node has one activation per pass through the same fork, so the most recent activation of a fork is only the asking scope's while that scope is the one running - once it is over, the newest belongs to a later pass.

A finished scope therefore read a running one's fork. #outputFor: asks whether the loop is over, was told "no" by another loop, and answered the slot of its own last iteration instead of its first value: the same activation said one thing while the second pass ran and another once everything was over.

The scope resolves its fork once, on the first use - which is the first step of its own scope, when the newest activation of that fork is still its own, since a second pass cannot start before the scope is popped - and keeps it. Every other node is looked up as before; for those the most recent one is right, because only a live scope looks a node up and its own activations are the newest there are.

Found while tracing a customer report of 2026-09-18 (a loop that ran twice wrote its first pass's last element 300 times); the reading path that report went through is ApptiveScript's step(id), fixed separately in apptive-base.