Scenario lessons fail when they read as war stories: entertaining, ungradeable, and useless under pressure. The answer-critique shape fixes that: each question gets a strong answer, a weak answer, why the scores differ, and a rubric tally. This lesson applies the shape to three short scenarios, then collects the five lessons the two drills teach. Rehearse these aloud; they are the closest thing in the section to the actual interview.
Weak answer: the ROM must be corrupt, so reprogram it and check the tools. It guesses one cause, names no check order, and would reprogram a ROM that was never read. Strong answer: check clocks and reset release first (53.948 us sequence, CPU last), then the ROM select equation, then strap sampling, then fabric outstanding transactions, then debug access; bisect with simulation versus emulation. The strong answer scores 3s because it is an ordered plan with a check per step; the weak answer scores 1s because it is a single guess. Lesson: triage is a sequence, and the sequence is the answer.
Weak answer: the DMA is broken, swap in a new one. It discards the distinguishing evidence: COUNT right means the engine moved exactly the programmed words, so the engine works and something changed the data, the key, or the source. Strong answer: read back KEY and LEN (writes are silently ignored while BUSY), check that the SBR and DMA lengths agree, check the slot alias, then the secure attribute on the path. Lesson: let the counters exonerate before suspecting the engine; correct counts with wrong data always mean the data path, never the control path.
Weak answer: waive them all, the schedule matters. It confuses urgency with evidence and would sign any risk. Strong answer: classify each waiver by blast radius and evidence (equivalence clean, timing with margin, CDC structural with a test), name the silicon fix for each (metal fix, software workaround, none), and state the decision rule: no waiver without a fix path ships. Lesson: signoff is risk classification with a fix path per item, and the decision rule is stated before the items, never after.
One: order beats cleverness; the triage rule outscores the lucky guess. Two: one observation per diagnosis; COUNT right, RAW bit 6, SLVERR on the first beat each name exactly one bug. Three: every fix ships with its audit; the reset-order assertion, the overlap check, the isolation count of 114. Four: budgets are subtractions, not adjectives; 2.74 mW over is a decision, substantially over is a meeting. Five: state the decision rule first; who owns the margin, what ships without a fix path, what placement wins before hardware.
The Evidence part: every step needs its verification beside it (assertion, directed test, audit count). Add one check per step and the answer gains two to three points without changing its content.