Why does a large "scroll" query fail with "circuit_breaking_exception" yet "reindex" API succeeds?

Question

I'm testing the ElasticSearch connector for Presto and running into an issue with queries that return a large number of documents. These queries fail due to a circuit_breaking_exception. For example, a query that filters on a single field (accountId stored as a number) and returns four high cardinality fields fails in this manner. After reading the docs, my understanding is that this exception is related to a node running out of memory and ES preemptively stopping the query to prevent an OOM exception.

Beneath the hood, the ElasticSearchConnector is using scroll queries.

Recently, a colleague of mine reindexed this account using the reindex API . My understanding is that this API uses a combination of scroll queries and the batch API to perform reindexing.

If my current understanding is correct, I'm confused as to why the reindex API would succeed while a vanilla scroll query would fail.

I understand that I could tinker with the memory available to the ES nodes in conjunction with increasing the circuit breaker memory limit, but before I go throwing hardware at the problem I'd like to make sure I'm not missing some fundamental difference between these APIs that causes one to succeed and the other to fail.