Thank you for your reply. I have carefully explored the mechanism and believe the role of coverage is as follows:
When new coverage is observed, the corresponding program (prog) is added to corpus.db.
The fuzzer itself is implemented independently of coverage, but through the selection mechanism, coverage feedback is provided to the fuzzer. This means that programs (seeds) that generate new coverage are given higher priority for mutation and execution.
I’m not sure whether my understanding is correct. May I ask if you have any further comments or additions? I look forward to your reply.