Checking that their product is being built with Kafka or whatever is worthless - you cam throw money at that problem later and fix it.
Your second paragraph is spot on, and I hinted at a similar idea in a sibling to your comment. Especially at series A and later, customers are an existence proof for the product not being vaporware.
So when we are involved in diligence, it’s usually near the end, and we’re getting to know the engineering leaders to see if we can help them speedrun a few iterations by sharing best practices or introducing them to other leaders in the portfolio who have been there and done that recently.