sometimes I also feel it tries to optimise for "per line coverage" over more "real, complex use cases" type tests