I doubt that it is as simple as that. Designing modern technology is incredibly hard, then running that design through a process that makes it viable for mass production is almost as hard. In both of those steps there might be tons of reasons to have it in one place or another.
Imagine the following scenarios:
- It has been found that during assembly units that do not pass QA often has this chip fail, and thus making it easily swappable increases the production yield during assembly instead of PCBA.
- To create a security boundary for biometric data, as a rule the assembly that captures the data and authenticates it has to be separate from the application processor assembly. Swapping either then cannot be done surreptitiously. That makes the chip's job part of the screen assembly and not something else.
- A cost reducing measure between the sensor and the data processing chip over a high-speed serial connection turned out to be removal of buffers or transceivers but this meant that the signal lines had to be shorter and sandwiched against a large ground plane to ensure signal integrity and prevent frequent retransmissions.
"Hostile" engineering (which is not a word I would choose) would be any kind of security screw that doesn't improve over other screws. Think of normal Torx (which is excellent) vs. "security" Torx, or Apple's Pentalobe. Those exist for one reason: to deter the average user from undoing those screws.
The reason for deterrence vary, in some cases it's just a reduction of liability ("we made it harder so that Joe Bloggs has to be really sure he wants to screw up his device, therefore we are not liable for his screwed up device"), sometimes theft ("we'll use security Torx screws on those coat hangers in the toilet stalls, that way the teens don't steal them to sell them on the black market"), sometimes it's some weird protection scheme that never works ("if we make the screws just a tiny bit harder to find a bit for, they will have to come to us and we'll have to deal with all of their problems instead").
In most mass-market cases it generally boils down to liability and scalability. And waiving liability as a middle ground often doesn't exist because it doesn't scale well.
What does help is making it part of a brand's identity or marketing, that way the budget doesn't have to come out of any of the general product lifecycle budgets which then makes it seem to have the same bottom line which is the whole reason for-profit companies exist.