Remember that with iOS devices, Apple is the chipmaker.
According to Apple, they designed the chips (CPU) inside the iPhone. And if uses licensed design components from ARM. But the chips inside the iPhone are not designed and produced by ARM, they are designed and produced by Apple, who uses some design elements from ARM under licensed permission.
So it is totally within Apple's capability and responsibility to modify the design of the chips used in iPhones and fix the flaw before sending out and additional defective parts to the wild.
To think they'll stop selling everything until they can do what neither ARM nor Intel have apparently done yet - come up with a new chip design that doesn't have those problems - is absurd. A similar situation applies with other devices that use ARM chips; ARM doesn't make chips, they license the design, but only huge, rich Apple is expected to make crushing sacrifices to avoid selling unfixed chips? You're kidding, right? And that would have violated the term of confidentiality for the vulnerability, presumably agreed to by others than just Apple.
Moreover, the terms of delayed vulnerability disclosures are clear; no one recipient should release info early. Such arrangements include a deadline by which the info will be released; but until then, all that might be able to work on fixes or mitigations need the time to do the work; and esp. with vulnerabilities as difficult to exploit as Spectre (the one that affects ARM), while the announcement is delayed, it's safer to keep the secret, because it's doubtful someone else will discover it.
These actually got out a few days early, because various OS vendors were spotted making somewhat similar changes for Meltdown. The most pressing work having been done, and updates already being pushed out, slightly early disclosure was better than rumors.
Intel deserves some market punishment (not necessarily lawsuits!) for Meltdown, IMO; but Spectre (which is also said to affect AMD, ARM, Power, even zSeries mainframes, and perhaps others) is complicated enough, and different enough in the details of exploitation on every CPU model, let alone different makes, that singling out any one entity that has particular responsibility for not having fixed the problem, seems like a can of worms simply not worth opening. Aside from possible indications that Intel has been aware of the general class of problem long before the papers on Meltdown and Spectre were written, IMO the process has worked largely as it should. Advanced designs yielded performance; advanced designs are almost NEVER provably correct, let alone provably correct to a standard that exceeds the formal specifications. Later, researchers discovered the problems, disclosed them quietly to manufacturers and vendors, who worked on getting fixes out for the problem before the public disclosure deadline.
