A global threshold of 80% says nothing about whether the code that matters is tested, and it is satisfied by testing the code that is easy.
<coverage>
<report>
<clover outputFile="coverage.xml"/>
</report>
</coverage>
<!-- and in CI, per path -->
src/Domain/Pricing min 95% actual 97%
src/Domain/Billing min 95% actual 96%
src/Http min 60% actual 71%
src/Infrastructure no floor
<!-- global: 74%, and nobody looks at it -->
Different code deserves different confidence, and one number averages that away. The pricing directory at 95% is a deliberate statement about where a bug is expensive; the infrastructure directory with no floor is an equally deliberate statement that its tests are integration tests measured elsewhere. The global figure went down after this and the suite got better.