Crawl Same-Hostname URLs Concurrently

Crawl the finite web graph described by pageUrls and linkLists, beginning at startUrl. The string linkLists[i] contains the absolute URLs returned when pageUrls[i] is parsed, separated by single spaces; an empty string means no links. A normalized URL absent from pageUrls has no outgoing links.
Before any comparison or lookup, remove the first # and everything after it. Visit only normalized URLs whose hostname exactly equals the normalized start URL's hostname. Different schemes may still have the same hostname. Never visit a normalized URL more than once, including through fragments or cycles.
Use exactly workerCount workers for the crawl. Return every visited normalized URL in lexicographic order.

Function
crawlSameHostname(startUrl: String, pageUrls: String[], linkLists: String[], workerCount: int) → String[]

Examples
Example 1
startUrl = "https://docs.acme.com/root#intro"
pageUrls = ["https://docs.acme.com/root","https://docs.acme.com/a","https://docs.acme.com/b"]
linkLists = ["https://docs.acme.com/a#one https://docs.acme.com/a#two https://other.com/out","https://docs.acme.com/b#part https://docs.acme.com/root#back","https://docs.acme.com/a"]
workerCount = 4
return = ["https://docs.acme.com/a","https://docs.acme.com/b","https://docs.acme.com/root"]
Both fragments of /a normalize to one URL. The cycle through /root and /a cannot create another visit, and other.com is excluded.
Example 2
startUrl = "https://api.site.com/home"
pageUrls = ["https://api.site.com/home","http://api.site.com/v1"]
linkLists = ["http://api.site.com/v1 https://api.site.com/unlisted#top","https://api.site.com/home#again"]
workerCount = 2
return = ["http://api.site.com/v1","https://api.site.com/home","https://api.site.com/unlisted"]
The scheme does not affect hostname equality. The unlisted URL is visited but contributes no outgoing links.
Example 3
startUrl = "https://solo.example.com"
pageUrls = ["https://known.example.com/page"]
linkLists = ["https://solo.example.com/hidden"]
workerCount = 1
return = ["https://solo.example.com"]
An unlisted start page is still visited. Its parser result is empty, so the crawl ends immediately.

Constraints
1 <= pageUrls.length == linkLists.length <= 120.
The input contains at most 480 links in total.
1 <= workerCount <= 4.
Every URL is an absolute http or https URL of at most 300 characters, contains no whitespace, and has a lowercase ASCII hostname with no credentials or explicit port.
Entries of pageUrls are fragment-free and unique. Paths and query strings remain case-sensitive; only fragments are removed.

It reduces to BFS or DFS over a map from URL to its link list. Normalize first: cut at the first # and drop the rest. Parse the hostname as the text between :// and the next /, ?, or end of string. Compare hostnames only, never schemes. Build a dict from pageUrls to split linkLists, and use an empty list for any URL not in the dict. Add a URL to visited when you enqueue it, not when you pop it, so cycles and duplicate fragments can't double-visit. Sort at the end. The pitfalls: forgetting to normalize the start URL, treating an unlisted URL as unvisitable, and splitting an empty string into one empty token. Filter empties. If you do implement real threads, guard the visited set with a lock, but a sequential simulation returns the same answer. StealthCoder is the hedge if the hostname parsing trips you up live.

StealthCoder is the hedge for the one pattern you didn't drill. It runs invisibly during the screen share.