Your second example disproves your point. In a bank transaction if anything goes wrong the transaction should abort. It is precisely when you cover the error that money can be lost or duplicated. Checking for null would be like "The customer sends null money... well I'll just send $10 (the equivalent of the image of the developer)"

I have updated the question with additional details.

In out case it was not the argument that was null. We made a database call for something that was expected to be present but in practice it was not present. The question is if we should check that every object exists or we should check only if we expect it to be missing.

Well it needn't be this specific article. What is more what is wrong with gaining knowledge by reading an article? This question demonstrates some understanding of multithreading. The series of articles introduces the async feature in C# 5.0. I guess asynchronous programming is often related to multithreding. And finally I do not suggest these 2 as the only questions :)