Memory exhausted from initial entity type query

The 2 initial ideas that come to mind are:

1- Instead of getting all revisions at once in the query:

    if ($entity_type->isRevisionable()) {
      $revision_key = $entity_type->getKey('revision');
      $query = $storage
        ->getQuery()
        ->allRevisions()
        ->sort($revision_key, 'ASC')
        ->accessCheck(FALSE);
      if ($multi_pass) {
        $query->condition($revision_key, $current_id, '>');
      }
      $result = $query->execute();
      foreach ($result as $revision_id => $id) {
        $queue->createItem([
          'entity_type' => $entity_type_id,
          'entity_revision_id' => $revision_id,
        ]);
        if ($multi_pass) {
          $multi_pass_from_state[$entity_type_id] = $revision_id;
          \Drupal::state()->set('entity_usage.multi_pass', $multi_pass_from_state);
        }
      }
    }

we could try to do that in separate loops, but still in the same PHP request/execution. For example we could set an "arbitrary" loop size, maybe 100, 500, 1000?, and then create an additional for loop that will keep querying for revisions, getting only this number each time, and then creating only this number of queue items in every loop. I'm not really sure it would avoid the memory issue but maybe it's worth trying.

2- Use a batch process to create queue items. This would be a more significant refactor, since it would mean to change the whole command to be executed in background, in batches. Here's an example of a drush command that is executed as a background batch process.

I would be reluctant to just do it in several "passes" without having a way to enforce everything is covered, since it could lead to some entities being missed (for example if they are created more or less at the same time). Now that I think of it, I'm kind of reaching the conclusion that the if ($multi_pass) approach might not have been the best idea in the end... :)

Committed, thanks for helping!