RMP.3:4.5 - Test the whole arrangement and choose the exact return
Read the completed arrangement as a possible piece of work. Could it still expose the error or rival that justified the selected design? Check interactions among selection, access, measurement, timing, resources and records, including their jointly admissible unresolved values.
RMP.2:4.3 supplies the joint-input condition: every admitted combination of PostSelectionNeed values preserves the selected family, core contrast, criticism, feasibility and refusal conditions. A PostSelectionNeed is therefore a bounded remaining need, not anything the team happens to postpone. Apply that condition to the actual combined specification. Do not test each uncertainty while holding the others artificially favourable.
| What the specification reveals | Action and result |
|---|---|
| The choices and their joint effects stay within the selected question/design bounds. | Complete or amend the protocol and its reasons. State the supported scope, remaining bounded needs and change conditions, then proceed only with the supported work. |
| A particular external result is still missing, but every admissible value preserves the selected design. | Name the needed result, its owner, the action it blocks and the condition for resuming that action. Retain the design and the usable portions of the specification; do not call the blocked part executable. |
| A new value can defeat the selected family, contrast, criticism, feasibility or refusal condition. | Return that value and its consequences to RMP.2. Preserve the previous decision and any actual work in the trace. The protocol is not a place to silently select another design. |
| The question, intended knowledge change, population/context claim or useful stop changes. | Return the changed question to RMP.1 and carry its answer into the dependent design and specification. |
| Permission, resources or another practical decision are withheld independently of evidence sufficiency. | Return to that responsible owner. Preserve the research question and identify the prohibited action. Inability to proceed is not an evidence-sufficient stop. |
The distinction can be small but decisive. Waiting for the exact calibration result can block a measurement while the design remains justified if all admitted results preserve the required discrimination. Discovering that the instrument may not distinguish the competing mechanisms reopens design selection. A numerical value is not “mere detail” simply because it arrives late.