Experiment with restarting secondary workers

[louismerlin]

As described in this post we could experiment with restarting secondary workers on an empty corpus to find new coverage.

For example, every 8th fuzzer could have the following lifecycle:

1. start on empty corpus without any shared seeds
2. run for T (e.g. T=1 hour)
3. sync corpus both ways
4. run for T
5. stop fuzzer and back to 1

[louismerlin]

@vanhauser-thc I'd love to hear your thoughts on this, if you think it would be a worthwhile experiments and if you'd change anything.

[vanhauser-thc]

sounds like a horrible idea :-)

you could try instead: split the corpus into X parts, and start X fuzzers without syncing between them, once X - 1 do not find coverage anymore merge their corpora and restart?