(crawler) Further rearrange crawl order

Limit crawl order preferrence to edu domains, to avoid hitting stuff like medium and wordpress with shotgun requests.

code/processes/crawling-process/java/nu/marginalia/crawl/CrawlerMain.java

@@ -302,8 +302,8 @@ public class CrawlerMain extends ProcessMainClass {
     }
 
     /** Create a comparator that sorts the crawl specs in a way that is beneficial for the crawl,
-     * we want to enqueue domains that have common top domains first, but otherwise have a random
+     * we want to enqueue domains that tend ro be large and have common top domains first,
-     * order.
+     * but otherwise have a random order.
      * <p></p>
      * Note, we can't use hash codes for randomization as it is not desirable to have the same order
      * every time the process is restarted (and CrawlSpecRecord is a record, which defines equals and
@@ -311,15 +311,13 @@ public class CrawlerMain extends ProcessMainClass {
      * */
     private Comparator<CrawlSpecRecord> crawlSpecArrangement(List<CrawlSpecRecord> records) {
         Random r = new Random();
-        Map<String, Integer> topDomainCounts = new HashMap<>(4 + (int) Math.sqrt(records.size()));
         Map<String, Integer> randomOrder = new HashMap<>(records.size());
 
         for (var spec : records) {
-            topDomainCounts.merge(EdgeDomain.getTopDomain(spec.domain), 1, Integer::sum);
             randomOrder.put(spec.domain, r.nextInt());
         }
 
-        return Comparator.comparing((CrawlSpecRecord spec) -> topDomainCounts.getOrDefault(EdgeDomain.getTopDomain(spec.domain), 0) >= 8)
+        return Comparator.comparing((CrawlSpecRecord spec) -> spec.domain.contains(".edu"))
                 .reversed()
                 .thenComparing(spec -> randomOrder.get(spec.domain))
                 .thenComparing(Record::hashCode); // non-deterministic tie-breaker to

code/processes/crawling-process/java/nu/marginalia/crawl/retreival/CrawlDelayTimer.java

@@ -3,6 +3,7 @@ package nu.marginalia.crawl.retreival;
 import nu.marginalia.crawl.fetcher.HttpFetcherImpl;
 
 import java.time.Duration;
+import java.util.concurrent.ThreadLocalRandom;
 
 import static java.lang.Math.max;
 import static java.lang.Math.min;
@@ -53,12 +54,13 @@ public class CrawlDelayTimer {
     public void waitFetchDelay(long spentTime) {
         long sleepTime = delayTime;
 
+        long jitter = ThreadLocalRandom.current().nextLong(0, 150);
         try {
             if (sleepTime >= 1) {
                 if (spentTime > sleepTime)
                     return;
 
-                Thread.sleep(min(sleepTime - spentTime, 5000));
+                Thread.sleep(min(sleepTime - spentTime, 5000) + jitter);
             } else {
                 // When no crawl delay is specified, lean toward twice the fetch+process time,
                 // within sane limits. This means slower servers get slower crawling, and faster
@@ -71,12 +73,12 @@ public class CrawlDelayTimer {
                 if (spentTime > sleepTime)
                     return;
 
-                Thread.sleep(sleepTime - spentTime);
+                Thread.sleep(sleepTime - spentTime + jitter);
             }
 
             if (slowDown) {
                 // Additional delay when the server is signalling it wants slower requests
-                Thread.sleep(DEFAULT_CRAWL_DELAY_MIN_MS);
+                Thread.sleep(DEFAULT_CRAWL_DELAY_MIN_MS + jitter);
             }
         }
         catch (InterruptedException e) {