Golang Channels and Deadlocks

GitLab's CI runner coordinator dispatches jobs to a pool of worker goroutines over one channel and collects their results over another. Under load, the whole coordinator occasionally freezes and never returns, and it always seems to happen when there are more jobs than workers.

package runner

func DispatchJobs(jobs []Job, workerCount int) []Result {
    jobChan := make(chan Job)
    resultChan := make(chan Result)

    for w := 0; w < workerCount; w++ {
        go worker(jobChan, resultChan)
    }

    for _, job := range jobs {
        jobChan <- job
    }
    close(jobChan)

    var results []Result
    for i := 0; i < len(jobs); i++ {
        results = append(results, <-resultChan)
    }

    return results
}

func worker(jobChan <-chan Job, resultChan chan<- Result) {
    for job := range jobChan {
        result := runJob(job)
        resultChan <- result
    }
}

Explain how the channels in this dispatcher interact, and walk through the specific sequence of sends and receives that leads to a deadlock once there are more jobs than workers. Then describe how you would redesign it to avoid that.