A Large Change Is Hard To Judge So It Gets Approved Unread

Beyond a certain size, review stops being an inspection and becomes a formality.

The uncomfortable dynamic is this. A reviewer facing a very large change has three options, and only one of them is honest. They can read it properly, which takes hours they do not have and which they know will delay a colleague. They can reject it and ask for it to be split, which feels obstructive after the work is already done. Or they can skim, form a general impression, and approve. The third option is quick, socially comfortable, and produces exactly the same record as a genuine review. Nothing in the process distinguishes them afterwards. The change is marked as reviewed, and everyone proceeds believing an inspection occurred.

This means that large changes carry a specific and counterintuitive risk: they receive less scrutiny than small ones, not more, despite affecting more of the system. The team's confidence in the review record is highest exactly where the record is least reliable. Nobody intends this and nobody is behaving badly. It is a straightforward consequence of finite attention meeting an unbounded demand, and it repeats wherever the size of the unit is left entirely to the discretion of the person who wants it approved. Worse, the effect compounds: a team that has approved several large changes without incident learns to trust the pattern, and the next one arrives with even less resistance than the last.

The fix is not to exhort reviewers to try harder, because the constraint is real and exhortation does not create hours. The fix is upstream, in how work is proposed. Ask for the split before the work starts, when it costs a conversation rather than a rewrite. Treat a request to break something up as a normal part of planning rather than a criticism of the author. And accept, as a team, that a reviewer saying they cannot properly assess something at this size is doing their job rather than avoiding it. That sentence needs to be safe to say, or it will not be said.