About one decade ago, the hybridized discontinuous Galerkin method was introduced in, which distinguishes itself by several unique features from other DG methods. For large-scale computations, however, hybridization alone has often turned out insufficient to overcome memory and time-to-solution limitations, resulting in ongoing algorithms-centered research. We investigate a finite element discretization strategy that combines elements of the continuous and discontinuous approach. Its main idea is to apply the hybridized DG concept not between individual elements, but between larger macroelements that can contain a flexible number of finite elements themselves. The regular simplicial patches have non-matching interfaces, enabling adaptive local refinement by uniform simplicial subdivision. This methodology towards local adaptive refinement, domain decomposition and load balancing does not require intermediate recourse to complex re-meshing tools, nor sophisticated load balancing procedures. Furthermore, the approach scales efficiently to n-node clusters and can be tailored to available hardware by adjusting the local problem size to the capacity of a single node. Increasing the local problem size means simultaneously decreasing, in relative terms, the global problem size. While the former is solved using direct solvers, the latter is solved using distributed linear iterative techniques. We study the efficiency, scalability, performance and load balancing of the described method in the context of several two and three-dimensional linear advection-diffusion problems. The macroelement HDG strategy leads to savings in both speed and memory, and enables excellent scalability for parallel computations. In addition, the macroelement HDG discretization mitigates the main disadvantage of standard discontinuous Galerkin methods, i.e. the proliferation of degrees of freedom, while maintaining many of their favorable numerical properties.