Mixed-model assembly lines are state of the art in automotive production systems. Because of the high number of customizable options which can be ordered in a vehicle, there is a huge variety of possible products. An important problem in this context is the sequencing of such products. Inevitably, there will be deviations from the intended production sequence in the course of production, as disruptions occur. The products must then be resequenced to ensure an optimal sequence. In this work, we consider the usage of a buffer (in form of an automated storage and retrieval system) between the paint shop and the final assembly to resequence the orders. We consider a high number of variants and, with this, a random input sequence for the buffer. Additionally to the physical resequencing in the buffer, the options get decoupled from the products. That allows virtual resequencing, in which parts and materials are interchanged. The dispatching selection must be made without full information in an online problem. To solve this problem, different heuristics and a lookahead algorithm are applied to minimize the amount of utility work in a paced automotive assembly line.