| "I'll just rewrite this from scratch — it'll be faster." | Throwing away tested, battle-hardened code throws away every bug fix and edge case it absorbed. The rewrite will rediscover those bugs the slow way. (97/6) |
| "There are no tests, but the change is obvious." | "Obvious" is how production breakages are born. Pin behavior with a characterization test first, then refactor. (97/6) |
| "I'll do it all in one big PR — easier to review." | Big PRs hide bugs and frustrate reviewers. Many small commits keep tests green and changes reviewable. (97/6, 97/24) |
| "The old code is ugly and uses outdated patterns — I should modernize it." | Style is not a refactor goal. New framework / new language / personal preference are not valid reasons. State the actual user-visible benefit or stop. (97/6) |
| "It's just a small cleanup, no need for a checklist." | The small cleanups are exactly where the worst tangles hide. Run the checklist. (97/74) |
| "I'll fix everything I see while I'm in there." | Boy Scout rule says a little better, not perfect. Bounded improvement only. Write the rest down for later. (97/8) |
| "I can patch it directly on the staging/production server, just this once." | No. Refactors flow through your normal commit → test → review → deploy path. "Just this once" is how outages happen. (97/31) |
| "The tests are failing but it's just flaky — I'll keep going." | Failing tests during a refactor mean the refactor changed behavior. Stop, investigate, fix or revert. Don't push through. (97/6) |
| "Estimating is too hard — I'll figure it out as I go." | Open-ended refactors balloon. Identify the coupling hotspots up front and re-estimate. If it's now bigger than the ask, escalate. (97/74) |
| "I'll just rename some variables — the function isn't that long." | If the function scrolls, renaming alone won't help. Extract helpers whose names explain the why; the body shrinks to a sequence of named steps. (Fowler/LongMethod) |
| "The function does X and also Y, but they're related." | If you needed "and also" to describe the unit, it has more than one reason to change — the SRP refactoring trigger. State the unit's responsibility in one sentence; if you can't without "and also," split. See clean-code decision 4 for the discipline of writing the result. (97/76) |
| "The same change keeps forcing me to edit the same eight files." | Shotgun surgery: the behavior is conceptually one thing, physically scattered. Move related fields and methods together until the next instance of the change is one file. (Fowler/ShotgunSurgery) |
| "This function already takes seven primitives — I'll just add an eighth." | Data clump. The fields are a missing type. Extract a class / dataclass / parameter object before adding the eighth. (Fowler/DataClumps) |
| "Everything lives in one file — it's easier to find when it's all together." | Co-location by import convenience is not a single responsibility. If the pieces change for different reasons, split the module along the axes of change. (97/76) |
