How should I select working memory tasks to balance validity and confounds?

[Frank_J]

I'm a graduate student in cognitive psychology researching the effects of chronic stress on working memory capacity in adults, and I'm currently designing my experimental protocol. I'm trying to decide between using classic n-back tasks, complex span tasks like the operation span, or more ecologically valid simulations for measuring real-world working memory load. For researchers in this field, what are the current best practices for task selection to maximize both internal validity and practical significance? I'm particularly concerned about practice effects and task-switching costs when administering a battery of tests, and I'd appreciate insights on reliable software packages for presentation and data collection, as well as how you effectively control for individual differences in baseline anxiety or sleep quality that might confound the stress manipulation.