Bug when starting from a large corpus

[louismerlin]

When starting a fuzzer from a large corpus with many cores, and the input corpus is in the default corpus directory, the fuzzers have a great chance of stopping with the "short read" error message.

This does not happen anymore when I use the -i parameter to select another initial corpus directory.