*
   * <p>Note: This default implementation is overkill (but valid) for an unordered container. An
   * equally valid implementation for an unordered container is to throw an exception. The chosen
   * implementation, however, has the advantage of working for insertion-ordered containers, as
   * well.
   */
  @Override
  public List<String> order(List<String> insertionOrder) {
    return insertionOrder;
  }

  private void countRemaps(long h, AtomicLongMap<Integer> map) {
    int last = 0;
    for (int shards = 2; shards <= MAX_SHARDS; shards++) {
      int chosen = Hashing.consistentHash(h, shards);
      if (chosen != last) {
        map.incrementAndGet(shards);
        last = chosen;
      }
    }
  }

  private static final int ITERS = 10000;
  private static final int MAX_SHARDS = 500;

   * note that listeners added after this point may be executed before those previously added, and
   * note that the execution order of all listeners is ultimately chosen by the implementations of
   * the supplied executors.
   *
   * <p>This method is idempotent. Calling it several times in parallel is semantically equivalent
   * to calling it exactly once.
   *

   *       storedPermits, we serve from those first
   *   <li>When _unused_, we go right at a constant rate! The rate at which we move to the right is
   *       chosen as maxPermits / warmupPeriod. This ensures that the time it takes to go from 0 to
   *       maxPermits is equal to warmupPeriod.
   *   <li>When _used_, the time it takes, as explained in the introductory class note, is equal to

   * is iterated are the individual lists created, and these are not retained after iteration.
   *
   * @param lists the lists to choose elements from, in the order that the elements chosen from
   *     those lists should appear in the resulting lists
   * @param <B> any common base class shared by all axes (often just {@link Object})

   */
  /*
   * Just as Converter could implement `Function<@Nullable A, @Nullable B>` instead of `Function<A,
   * B>`, convertAll could accept and return iterables with nullable element types. In both cases,
   * we've chosen to instead use a signature that benefits existing users -- and is still safe.
   *
   * For convertAll, I haven't looked as closely at *how* much existing users benefit, so we should

   * and accesses that were performed on the map. The queue is drained on writes and when it exceeds
   * its capacity threshold.
   *
   * The Least Recently Used page replacement algorithm was chosen due to its simplicity, high hit
   * rate, and ability to be implemented with O(1) time complexity. The initial LRU implementation
   * operates per-segment rather than globally for increased implementation simplicity. We expect

     *       be achieved, for example, by repeating in the {@code META-INF/services/} directory the services
     *       that are declared in the {@code module-info.class} file. In that case, the path type can be chosen
     *       by the plugin.</li>
     *   <li>If a {@link #patchModule(String)} is also set and the main JAR file is placed on the module-path,

   * <p>For the case of {@link RoundingMode#HALF_EVEN}, this implementation uses the IEEE 754
   * default rounding mode: if the two nearest representable values are equally near, the one with
   * the least significant bit zero is chosen. (In such cases, both of the nearest representable
   * values are even integers; this method returns the one that is a multiple of a greater power of
   * two.)
   *

   */
  /*
   * Just as Converter could implement `Function<@Nullable A, @Nullable B>` instead of `Function<A,
   * B>`, convertAll could accept and return iterables with nullable element types. In both cases,
   * we've chosen to instead use a signature that benefits existing users -- and is still safe.
   *
   * For convertAll, I haven't looked as closely at *how* much existing users benefit, so we should