Tools like strace can help you analyze a program's behavior from the outside, but you get limited insight into its internals (e.g., what algorithms is it using?).
Being open source does not automatically make software more secure. A successful compilation doesn't automatically make your code bug-free. Yet both are necessary to achieve the desired goal: security and correctness, respectively.
Or, do you trust that someone has looked at it? How much faith do you have in someone out there in the community?
My point isn't that open source isn't a good thing--my point is that it's not the silver bullet a lot of people blindly assume it to be. Hence the second line of my post:
> In theory it means something could maybe be safer, but it far from guarantees it.
I have debugged all sorts of issues with closed source applications using these tools.
The commented I replied to implied that one would walk through the entire stack, every line of code, to audit e.g. an app running on a phone. This is most certainly not what OP meant. Rather, on a whole OSS is mostly transparent, while proprietary software is not. There are of course bugs, but that's not the focus here with "safety".
What we care about is intention. Private companies's have a track record in implementing features that go directly against the benefit if their end-users, e.g. tracking or vendor-lock-in. These anti-features, like the one described in the article, are much harder to detect precisely because the software is proprietary.
In theory it means something could maybe be safer, but it far from guarantees it.