From ​“Separability of tight and roaming pathways to molecular decomposition” by L.B. Harding, S.J. Klippenstein and A.W. Jasper in the Journal of Physical Chemistry [A 116, 6967 (2012)]:

In roaming radical reactions, one bond in a molecule becomes greatly extended and, before the nascent radicals completely separate, they reorient and react, leading to unexpected products.

The transition state regions associated with roaming are characterized by loose low-frequency motions. Roaming may compete with more conventional chemical processes involving tight (harmonic) transition states. Loose and tight processes require different theoretical approaches for calculating accurate rates.

When competing tight and roaming pathways connect the same sets of reactants and products, these two regions are actually parts of a single unified transition state.

In our paper, we showed that tight and roaming pathways may be treated as separable to a good approximation, and we identified features of the potential energy landscape that serve as mechanism dividers (second order saddle points or minimum energy points on conical intersections). Importantly, competing mechanisms may be dynamically separable even when the two regions are not energetically separable.

We provided methods for understanding this separability and for making quantitative kinetics predictions for systems with competing processes. Roaming pathways (and, more generally, competing tight and loose pathways) may be expected throughout chemistry, whenever radical-radical products are likely, and theoretical methods like the ones presented here are necessary to accurately predict kinetics and product branching for these systems.