Anything in there is repairable or replaceable unless the board burnt up in some sort of catastrophic event. The issue becomes a matter of time and cost, which is where cloning and replacement/reprogramming comes into play as it becomes cheaper to replace it rather than pay a tech 10-20-30 hours or more to reverse engineer, trace issues and repair it. If a new ECU costs in the neighborhood of $1000 you get maybe 10 hours of tech time until you reach the point that a new ECU is cheaper. If a used/good ecu can be had for $100-200 you get at most a couple hours (tech + business overhead expenses) which effectively limits board level repairs to relatively minor fixes. Such as replacing a few discrete components -if- the problem is obvious (transistors, resistors, etc, or maybe one packaged SOC if the data/program code is available or can be read/transferred from the old chip easily.
If one has the expertise and can do this themselves at home without paying somebody else + business overhead then sure, spend two weeks figuring it out and fixing it if you want. But X can't do that if they expect to stay in business.
RichHale, thank you for correcting me on the specifics of how that CP driver works. I never dealt with a failed ECU coil pack driver circuit, thus I was assuming it operated more like an electronically controlled magneto (where current is dead, applied momentarily via completing the ground, the spark is generated at the end of the pulse width as the field collapses). I wasn't aware that the current stays on and is merely pulsed off to collapse the field/generate the spark. The method I envisioned (LED across the ECU to ground) would still work, except blink on instead of blink off, much like the $10 tool that probes across the coil pack instead of the ECU.
Since I probably (I know I do if I 'dig') have a suitable LED just floating around somewhere I would more than likely save my $10. Most people would buy the tool, however.