I use PySnooper[1] when code behavior deviates inscrutably from my mental model. Like Icecream, PySnooper exhorts, "Never use print for debugging again." A simple example:

    import pysnooper

    @pysnooper.snoop()
    def add_up(numbers):
        total = 0
        for number in numbers:
            total += number
        return total

    add_up([123, 456])

When run, PySnooper prints the activity of the decorated function or method:

        $ python example.py
        Source path:... /home/joel/example.py
        Starting var:.. numbers = [123, 456]
        08:46:58.742543 call         4 def add_up(numbers):
        08:46:58.742692 line         5     total = 0
        New var:....... total = 0
        08:46:58.742722 line         6     for number in numbers:
        New var:....... number = 123
        08:46:58.742755 line         7         total += number
        Modified var:.. total = 123
        08:46:58.742787 line         6     for number in numbers:
        Modified var:.. number = 456
        08:46:58.742818 line         7         total += number
        Modified var:.. total = 579
        08:46:58.742847 line         6     for number in numbers:
        08:46:58.742876 line         8     return total
        08:46:58.742898 return       8     return total
        Return value:.. 579
        Elapsed time: 00:00:00.000405

[1] https://github.com/cool-RR/PySnooper