The incremental analysis as it is assumes that globals are, at the point of the incremental preparation before the actual solve, either part of the contexts (no earlyglobs) or in the start states. The latter case occurs with earlyglobs and the older privatizations (none/protection-old/protection-vesal) because for these, globals are temporarily added to the start state before they are side-effected in sync.

The newer privatizations are however implemented such that globals are immediately side effected in the presolving step and are part of the start variables during the incremental preparations in TD3. The incremental analysis of the TD3 needs to be adapted to handle this appropriately. The following should be considered:

• set_start is executed for all start variables and overwrites the local state. This can be problem for globals in the start variables
• with configurations like ana.base.context.int globals are not necessarily part of the contexts