Naming

A name is a contract. Readers — above all diff reviewers, who see the call site but not the implementation — act on the name alone. A misleading name causes misuse; a vague name adds cognitive load (principle of least astonishment: code should do what its name makes you expect).

Flag, in order of severity:

• Misleading names (worst offense): the reader's natural assumption would be wrong.

  function sanitizeInput(s: string) { return s.trim(); }
  

  Every caller now believes injection attacks are handled. Likewise a get... that mutates, or a create... implying checks it doesn't do. Rename to what it really does (trimWhitespace) — or better, make the code honor the name's claim.
• Abbreviations and single letters: usrCnt, procDt(d). The reader must reverse-engineer your compression scheme. Write userCount, processDate(date) — autocomplete pays the typing cost, not you. (Idiomatic loop indices like i in a 3-line loop are fine.)
• Types encoded in names: Hungarian notation (strName), I-prefixed interfaces (unless the ecosystem convention, e.g. C#), and Base/Abstract in class names. Callers shouldn't care whether they hold an interface or a class. If the parent seems to need "Base" (BaseTruck), the child is under-specified: name the parent Truck and the child TrailerTruck.
• Missing units: delay — of what? delaySeconds at minimum; better, a unit-carrying type (Duration, TimeSpan) so the compiler makes the unit unmissable.
• Names that need a translator: if a comment's main job is to define the name — "override" meaning the lifting of a billing freeze — the name failed; rename (billingFreezeLiftRequested) so the comment can shrink or vanish. Watch for ceremony words — handle, process, resolve, manage, decide, -State, -Manager — which name plumbing instead of mechanism: manageAccountState says neither that it takes a row lock nor what the lock protects; lockAccountRow says both.
• utils/helpers/misc dumping grounds: a naming struggle usually means the code is in the wrong place. Move functions onto the types they operate on, or into cohesive modules that can be named (paging logic → Paginator, cookie parsing → Cookie). Standard libraries have no utils module for a reason.

Test every name: would a rushed reviewer, seeing only this name in a diff, make correct assumptions about its behavior, side effects, and ownership?