Apple's security paper says it would take more than 50 years to brute-force an alphanumeric 6-digit passcode at 80ms per iteration. I suspect that's still correct here (if “3 days or more” is for 6 numeric digits).
Of course that's non-trivial EE work, but the point is it's possible, for someone with enough money and the right equipment. What would make it intractable is to ditch the idea that a 4 digit pin is protecting you from anything. There's simply not enough entropy in that.
Time delays are useful protection when over a network. But not when the attacker has physical console access, e.g. to a phone. At that point proper cryptography and mathematics is the only good protection.
https://www.arm.com/products/security-on-arm/trustzone is the generic version of this, no idea what exactly Apple is doing.
Anyway, to break the encryption involves finding bugs in the software that runs in the secure zone (the same way you'd defeat a time delay in a networked application) or by opening up the CPU die and figuring it out with an electron microscope (perhaps while the CPU is running).
Ultimately, software is going to be the productive attack vector. While the TrustZone docs emphasize "small" for the amount of code that runs in the secure zone, no programmer can ever do "small" on a deadline and so I wouldn't trust it to provide any actual security. But the hardware infrastructure is there to be pretty secure against any adversary that doesn't have a lot of time or money, if the programmers do their job correctly. (I don't trust it because of the time schedules involved for these SoCs. One or two a year, with a lifetime of maybe a couple years. Nobody is going to spend the money to do the hard work like looking for bugs just so that someone can't rip Netflix video feeds. The engineering work would be more expensive than whatever their contract with the DRM vendors says they have to pay if they get hacked, which is the only incentive to be secure... and I doubt Apple would sign one of those.
Perhaps a nation-state actor could shave down the processor and read that key with a SEM or some crazy thing, but that's literally how far the design is supposed to have pushed iOS security. Which is what makes this hack so embarrassingly bad (if confirmed).
A security company can easily afford either.
Shaving down the processor is hard but still possible. A government with a billion-dollar budget can probably do it, if they want, including after practicing on a million throwaway phones in the process. Apple's design is also known as security by obscurity (https://en.wikipedia.org/wiki/Security_through_obscurity).
I much prefer an open design and security by high-entropy keys and well-established practices in the cryptography. I'd feel much safer using open cryptography libraries and strong passwords than arbitrarily and blindly trusting some secretive engineers at Apple.
Apple's design is super-hard to break but possible. My Linux laptop, on the other hand -- the hard drive is standard SATA. Have at it. I don't care if you have a billion dollars, you'll have a hard time breaking into my data for a probably at least a few decades.
The only reason your Linux laptop would be more secure than an iPhone is if you were using a high-entropy key to unlock it every time you wanted to use it -- and the iPhone wasn't. That's it. But remembering high entropy keys without storing them in some less secure manner is so inconvenient and failure-prone for most people that it's actually a less secure design than Apple's SEP-assisted approach. And you could opt-in to a high entropy key on an iPhone if you wanted to, so even that is a false comparison.
[1] https://images.apple.com/business/docs/iOS_Security_Guide.pd...
The SEP is supposed to enforce a time delay between passcode attempts to prevent this sort of brute forcing. The timer could be defeated in older models by cutting power at just the right time, but Apple's whitepaper says it's supposed to survive restarts now.[3]
Based on the screenshots it looks like it can load custom firmware on the iPhone. That's bad.
[1] https://www.anandtech.com/show/12525/security-researchers-pu...
[2] HN discussion: https://news.ycombinator.com/item?id=16597626
[3] p15: https://images.apple.com/business/docs/iOS_Security_Guide.pd...