| Treating the most recently received values as simultaneous current state | Retain sample times and propagate their constraints through actual intervening inputs and disturbances. |
| Selecting sensors solely by total estimation error | Calculate the actions, constraints and continuation that the remaining uncertainty permits. |
| Assuming a complete delayed observation removes uncertainty | Retain the disturbances accumulated since the sample; test whether any action can tolerate them. |
| Keeping only an estimate and losing the means to update it | Construct sufficient history or an equivalent recursive state; verify the next update after the proposed reduction. |
| Turning a smaller covariance or successful simulation into a hard bound | Use the stated stochastic conclusion or derive a separate bounded-set guarantee. |
| Calling a conservative test failure a physical impossibility | Refine the enclosure or provide indistinguishable histories with incompatible consequences. |
| Removing an internal path, keeping all gains, and concluding the path is universally necessary | Retune the competing architecture under the same requirement, or use an information-based impossibility proof. |
| Adding a second sensor without including shared transport and compute load | Recalculate timing and the resulting error under the joint realization. |