That it's difficult to accurately define "what the application thought it wrote" when considering corruption at various abstraction layers; somewhat similar to

That it's difficult to accurately define "what the application thought it wrote" when considering corruption at various abstraction layers; somewhat similar to calculating checksums over already corrupted data.