Insurance companies would be allowed to use medical underwriting to increase premiums for people with lapsed coverage in lieu of charging the normally-applicable surcharge, and those increased premiums would only apply for the same period during which the surcharge otherwise would've applied (generally, one full plan year). There is no mechanism for using underwriting for the policyholders' benefit. So this feature won't lead to separate, permanent risk pools.
I had written the statements quoted above based on my own reading of the text of the MacArthur amendment (a PDF of which is available
here). However, I just came across
this analyis (dated April 27th) by the Brookings Institution, which agrees with bacchi's assertion that the MacArthur amendment might effectively result in separate risk pools between which applicants can effectively choose (a recipe for death spirals).
The text of relevant provision of law that would be added by the MacArthur amendment (Section 2701(b)(1)(C) of the Public Health Service Act) says that one of the purposes for which a State may obtain a waiver is to "(i)...not apply any increase to the monthly premium rate that would otherwise apply under section 2710A
[i.e., the premium as increased by the 30% penalty]...and...(ii) instead, ... apply subsection (a)(1)
[i.e., the subsection that lays out the full universe of factors that can be included in determining the premium] as if health status were included as a factor described in...such subsection." Because of the manner in which this is drafted--with health-status-based underwriting specified as permissible "instead" of imposing the 30% penalty--I had read it to mean that insurers can only use medical underwriting to policyholders' detriment (in lieu of imposing the penalty), and not to their benefit. However, I am now persuaded that this is not the case.
However, because the lead-in to this waiver provision says "with respect to an individual who is an applicable policyholder
[i.e., a policyholder who can't demonstrate continuous coverage]...with respect to an enforcement period (as defined in section 2710A(b))
[i.e., the period during which the 30% penalty would ordinarily apply]," the waiver provision's scope appears to be limited to the period during which the 30% penalty would otherwise apply (generally, one plan year). So I still think insurance companies in opt-out States would only be allowed to charge underwritten premium rates on a temporary basis, and not on a permanent basis (in other words, the health status rating would only apply for the period during which the 30% penalty would have applied--which is never longer than one year). The Brookings Institution analysis seems to overlook this point.