I see that this is also a community edition of the server you're using. Could you try the exact same scenario on enterprise edition and let me know of the behavior you see?
So email@example.com+ creates a buffer for every connection for reuse - so there's a bit of allocation done at the start. This took the memory consumption by 5 connections up a bit and FTS' memory usage when it begins to ingest items takes it over the quota causing very slow ingest.
I just raised the FTS quota of your node to 1024MB, and restarted the same index - it completed very quickly.
So this comes back down to sizing - for this test, I'd recommend either ..
- raising the FTS quota if you want to create an index with a single partition, or
- partition your index for distributing it among all the nodes you have in the cluster.
And for why we didn't see logs from your initial test - i've no idea.