It's a trade-off. Larger input files may slow the fuzzing process, and therefore explore less of the problem space. You usually want to test many different kind

It's a trade-off. Larger input files may slow the fuzzing process, and therefore explore less of the problem space. You usually want to test many different kinds of inputs, not just more of the same. OTOH file formats often include sizes of fields, which a fuzzer will set to arbitrarily high values. This tests (some) handling of overly large inputs without files being actually that large.