I’ve run into this exact issue before. Your recursive approach is correct logically, but as you’ve noticed, it starts to struggle once nesting gets deep because of recursion overhead and call stack limits.

One practical way around this is to avoid recursion entirely and flatten iteratively using a stack. It tends to be faster and more predictable for unknown or deeply nested structures. Libraries like itertools are great, but they work best when the nesting depth is known or shallow.

If performance really matters and the data is large, I’ve also found it useful to question whether the structure needs to be fully flattened upfront, or if processing items lazily as they’re encountered is enough. That alone can save a lot of time and memory.

Curious what kind of data volumes you’re dealing with. That usually changes which approach makes the most sense.