Defending the take-home in the review call

The client stands, then walks a few steps across the shop in the trial shoe; the maker does not parade every stitch. Asked why the sole is leather, she answers from her order card: it breathes, and it slips on wet stone. When the client says the heel slips, she checks, says so, and adds a heel grip. When he would have liked brogue punching, she calls it taste, says what plain leather buys, and they settle. Asked for a padded tongue, she fits it, then has him walk again. Re-lasting the whole shoe waits for the next fitting.

At the fitting	In the review call
A few steps across the shop	A short run of the main path
The order card's reasons	The README's decisions
“The heel slips”: checked, a heel grip	A defect: good catch, the fix and its test
Brogue punching: taste, settled	A design preference: trade-off, settle
Walking again after the padded tongue	Every test run again after a live change
Re-lasting booked for the next fitting	A big change agreed as a follow-up

Breaks down when: the client's foot settles every question of fit, while a reviewer's preference is often one reasonable view among several. Wrong inference: “whatever the reviewer prefers is right.” A fit is also felt at once, while a code change shows only when its tests run.

In real terms: before you answer any comment, decide whether it names a defect or a preference.

EXH-2320 is a synthetic review call on the ExamHall proctor-dashboard take-home. At 25:00 of 45 the reviewer asks for “pause the exam for all candidates”. The candidate first argued that pausing was out of scope, then rewrote the countdown module as a class to add it “properly”, without running the suite first. MR !2620 is the re-run on the same request.

EXH-2320 · first call · request at 25:00 of 45: "pause the exam for all candidates"
  before the change: suite not run
  countdown rewritten as a class: 0 of 7 pass · 7 failing, each with: m.join is not a function
MR !2620 · re-run · same request at 25:00
  before the change: 6 of 6 pass
  the reviewer's case added as a test: 6 of 7 pass · failing: the reviewer's pause test (pauseAll is not exported)
  pauseAll and resumeAll added beside the existing functions: 7 of 7 pass

In the first call every test failed, because each one calls join, which the rewrite had turned into a method. With no count from before, she could not say whether the module had worked before she touched it. The argument about scope had used up time the change then needed.

In the re-run she says “six of six” before typing anything. She adds the reviewer's case as a test and shows it failing, which proves the test can see the missing pause. She adds two functions beside the existing ones, leaves join and extend as they were, and reports “seven of seven: the six old tests and the pause”.

The script counts tests on the author's fixtures; it does not judge the call or what a real reviewer would score.