{"id":"24b58527-4da2-4fa7-b71e-98f5d69193d0","org_id":"5b607c7a-3ba2-4077-b038-cdccd0ce7c34","project_id":"b18679c0-fffb-4308-82b6-9e91be23582f","parent_id":null,"identifier":"BEB-9227","number":9227,"title":"DYNAMIC BIND PROOF — can a worker resolve a declaration and load its implementation at call time? Four questions, one page, blocks BEB-9226","description":"Repo: alexrollin/abilifi-apps\nClass: CODE\nLane: part\nFEATURE NAME: DYNAMIC BIND PROOF\nBLOCKS BEB-9226. This is the biggest unknown in the declared-catalog design and everything rests on it.\n\n## Why this first\nThe whole catalog design assumes an intermediating worker can RESOLVE A DECLARATION AND LOAD ITS IMPLEMENTATION AT CALL TIME, with no per-tenant deploy. If that does not work today, the design is wrong in a way no amount of ticket writing will fix, and we need to know in a day rather than a month.\nThis is a PROBE, not a build. Smallest thing that answers the question.\n\n## The four questions, in order. Stop at the first NO and report it.\n1. CAN A WORKER LOAD AND RUN CODE IT DID NOT SHIP WITH, at call time, resolved from a declaration? One trivial frag, one call, real output. Yes or no with the mechanism named.\n2. WHAT IS THE CALL COST? Cold and warm, measured, not estimated. If a dynamically bound frag is an order slower than a deployed one, that changes the design rather than killing it — but it must be a number.\n3. DO FACETS COUNT AGAINST THE WORKER SCRIPT CAP (BEB-2868)? This is open since BEB-7515 and it decides how many tenants are affordable. If they count, per-tenant dynamic binding has a ceiling and we need to know where.\n4. WHAT CAN A DYNAMICALLY BOUND THING NOT DO that a deployed worker can — bindings, routes, custom domains, cron? Any of those missing shapes the catalog.\n\n## Read before building\nBEB-7515 asked most of these and its specs are SUPERSEDED as of 2026-08-31. FIND WHAT SUPERSEDED IT FIRST. If the answers already exist, this ticket is a lookup, not an experiment. Also read cloudflare-os (workshop-frontend, workshop-backend) — a published working version of this pattern exists and re-deriving it is the exact failure we just named.\n\n## Acceptance\nA one-page answer to all four questions, each with the evidence that produced it: the call transcript for question 1, measured timings for 2, a stated source or a real test for 3, and a named list for 4. A NO on question 1 is a complete and valuable result — report it immediately, do not work around it.\n\n## Explicitly unverified\nEverything above. That is the point of the ticket.\n\nRefs: BEB-9226 (blocked on this), BEB-7515 (superseded — find why), BEB-2868 (script cap), BEB-6847, doc 8bf1fded.","status":"todo","priority":0,"assignee_id":null,"created_at":1788346326443,"updated_at":1788443307110,"completed_at":null,"due_date":null,"claimed_by":null,"claimed_until":null,"required_capabilities":null,"estimate":200,"cost_ceiling":400,"cancelled_at":null,"last_actor":"unverified:shared"}