We try to solve it in a couple of ways:

* Enable smaller PRs. Reviewing 50-150 lines of code is much less daunting. This can be done by using tools like feature flags, AB tests, github/scientist etc.

* Ask/communicate about the quality of PRs. Like: splitting big ones into multiple,writing good descriptions, including the problem and screenshots.

* Automate the trivial parts. We use Pronto (disclaimer: I'm the author - https://github.com/mmozuras/pronto), but there are other tools/services that can help achieve the same result.