The author talks about high throughput achieved via batching. But he has not mentioned latency implications of batching, or perhaps I missed that in the article.

Wouldn't batching lead to an increase in transaction latency, even if we achieve higher throughput?