Pattern for sequential polecats with context isolation?

[HeadTacoDev]

I want to build a pipeline where polecat A implements a feature, then polecat B (fresh context) code-reviews that implementation without seeing how A reasoned about it. This is part of a larger sequential pipeline I'm designing, where context isolation between steps is a large point of it.

The sticking point: polecat A finishes its work on a branch but shouldn't merge — polecat B needs to review first. gt done --status COMPLETED submits to the merge queue, DEFERRED is semantically wrong. There doesn't seem to be a clean "I'm done with my step but this code isn't ready to merge yet" exit path.

Bead dependency chains feels like the correct solution here, but it seems to be running up against some fairly hard assumptions that 1 bead == 1 PR/unit of work. I also contemplated using sub-agents within Claude itself for this which isn't terrible, but in my experimenting outside of gastown still leaves a lot of room for phantom requirements to become real and other undesirable side effects. Is there an established pattern for "sequential fresh-context polecats sharing one branch, one merge at the end"?

I'm completely brand new to gastown so totally possible I've just missed a core concept (I'm still fairly fuzzy on exactly how formulas/molecules/proto molecules/etc. are all working together so maybe that's the gap?) somewhere along the way?