You need to use new channel for each Crawl. Each Crawl must be a go-routine (in the main function too). In order to crawl every url ONLY ONCE you need structure, with sync.Mutex and map[string]bool (i choose boolean type cause i do not found set data sctructure in standard libary). Also decision was inspired by https://stackoverflow.com/a/13223836/9475552 .

package main

import (
    "fmt"
    "sync"
)

type Fetcher interface {
    // Fetch returns the body of URL and
    // a slice of URLs found on that page.
    Fetch(url string) (body string, urls []string, err error)
}

type UrlCache struct {
    mu   sync.Mutex
    urls map[string]bool
}

func (uc *UrlCache) Add(key string) bool {
    uc.mu.Lock()
    defer uc.mu.Unlock()
    if _, ok := uc.urls[key]; ok {
        return true
    } else {
        uc.urls[key] = true
        return false
    }
}

var cache UrlCache = UrlCache{urls: make(map[string]bool)}

// Crawl uses fetcher to recursively crawl
// pages starting with url, to a maximum of depth.
func Crawl(url string, depth int, fetcher Fetcher, results chan string) {
    defer close(results)
    // TODO: Fetch URLs in parallel.
    // TODO: Don't fetch the same URL twice.
    // This implementation doesn't do either:
    if depth <= 0 {
        return
    }
    cache.Add(url)
    body, urls, err := fetcher.Fetch(url)
    if err != nil {
        results <- err.Error()
        return
    }
    results <- fmt.Sprintf("found: %s %q\n", url, body)
    temp_results := make(map[int]chan string)
    for i, u := range urls {
        if !cache.Add(u) {
            temp_results[i] = make(chan string)
            go Crawl(u, depth-1, fetcher, temp_results[i])
        }
    }

    for _, temp_res := range temp_results {
        for tr := range temp_res {
            results <- tr
        }
    }

    return
}

func main() {
    crawl_results := make(chan string)
    go Crawl("https://golang.org/", 4, fetcher, crawl_results)

    for r := range crawl_results {
        fmt.Println(r)
    }
}

// fakeFetcher is Fetcher that returns canned results.
type fakeFetcher map[string]*fakeResult

type fakeResult struct {
    body string
    urls []string
}

func (f fakeFetcher) Fetch(url string) (string, []string, error) {
    if res, ok := f[url]; ok {
        return res.body, res.urls, nil
    }
    return "", nil, fmt.Errorf("not found: %s", url)
}

// fetcher is a populated fakeFetcher.
var fetcher = fakeFetcher{
    "https://golang.org/": &fakeResult{
        "The Go Programming Language",
        []string{
            "https://golang.org/pkg/",
            "https://golang.org/cmd/",
        },
    },
    "https://golang.org/pkg/": &fakeResult{
        "Packages",
        []string{
            "https://golang.org/",
            "https://golang.org/cmd/",
            "https://golang.org/pkg/fmt/",
            "https://golang.org/pkg/os/",
        },
    },
    "https://golang.org/pkg/fmt/": &fakeResult{
        "Package fmt",
        []string{
            "https://golang.org/",
            "https://golang.org/pkg/",
        },
    },
    "https://golang.org/pkg/os/": &fakeResult{
        "Package os",
        []string{
            "https://golang.org/",
            "https://golang.org/pkg/",
        },
    },
}