The purpose of this paper is to address some theoretical issues related to the track-to-track fusion problem when the measurements tracking the same target are inherently correlated by the common process noise of the underlying target. This problem has been intensively investigated using standard Kalman filter with some appealing theoretical results, however such results are no longer valid in case of suboptimality due to either the presence of strong nonlinearity or to the discrete uncertainty pervading the origin of the measurement. This paper reviews several architectures of parallelized blocks of Kalman filters, including the augmented stacked measurement, sequential and data compression architectures. Next, convex combination architecture will be investigated and some theoretical results concerning its extension as well as in case of presence of correlation are investigated. Two special cases of correlation are highlighted. This concerns the case of presence of only two correlated tracks among all tracks and the case of weak correlation. In both cases some original theoretical results are put forward. Finally, links with related fusion architectures is investigated.