We now handle large minOccurs/maxOccurs on element/wildcard particles more gracefully by creating a compact representation in the DFA and using counters to check the occurence constraints, however we will still fully expand the content model for minOccurs/maxOccurs on sequences and choices which could still lead to an OutOfMemoryError or very poor performance (i.e. could still take several minutes to build the DFA). Sequences, choices and nested minOccurs/maxOccurs are somewhat tricker to handle. We would need a more general solution than the one implemented for elements and wildcards to improve those.
With the introduction of XML Schema 1.1 support we would also need to consider how to improve this for the enhanced xs:all model groups.