Consider we have big XML file and we can't load it all into memory. We use then `iterparse` function from XML.etree.ElementTree module to parse it element by element.

Problem is, XML doesn't allow to run this smoothly and starts outputing wrong data after loading 16 kb (16*1024, found it after looking into source code). Having large number of children, we get the information that we have just a few.

To reproduce the problem, I created this example program. It makes simple xml file with progressively bigger files and tracks how many children of main objects there are counted. For small objects we have actual number, 100 children. For bigger and bigger sizes we have smaller numbers, going down to just few.

This is not a bug, it's normal, documented behaviour. The children are not guaranteed to be available during the "start" event. Only the tag itself is guaranteed to be there. The guarantee that the subtree is complete is only given for the "end" event.

See the big note in the documentation:
https://docs.python.org/3/library/xml.etree.elementtree.html#xml.etree.ElementTree.iterparse