Suppose if we support Platinum over molecular sieves, it dehydrogenates hydrocarbons first followed by protonation, isomerization and hydrogenation. What makes the Platinum dehydrogenate and same metal will hydrogenate latter.
To my opinion radicals are not involved in these reactions. The ability of oxidative addition and reductive elimination of similar or the same molecules is an intrinsic and necessary property of many transition metals, which is necessary to drive the catalytic cycles. In your question you are probably referring to isomerization or hydroisomerization of alkanes. In a simple alkane isomerization with Pt on charcoal you don't even need to dehydrogenate and hydrogenate, the reaction can proceed through skeletal isomerization via several different intermediates. (see http://books.google.hu/books?id=KSrTF6ZmkrwC&pg=PA182&lpg=PA182&dq=isomerization+of+alkanes+platinum&source=bl&ots=ErlFKxvhOf&sig=uUDwgOff0wSAatrknqWd9LOuVzs&hl=en&sa=X&ei=PjzpU5iIGKSr0QW354EI&ved=0CLkBEOgBMBE ). In case of bifunctional catalyst, i.e. Pt on acidic support, the reaction (the exact mechanism is still debated) can go through a carbenium ion or an olefin formation depending on the acidity of the support. In former case the acid protonates the alkane to an unstable carbonium ion, which decomposes to carbenium ion and hydrogen. The surface bound carbenium ion isomerizes then and react with a Pt-H (hydrogen source). If the support not so acidic, the reaction can go through olefin formation, which can isomerize via acid- (carbenium ion formation) or via Pt-catalysis (Pt-alkyl formation in reaction with Pt-H). The product is obtained with a hydride in both cases.