There is a reason why little endian is preferred in virtually 100% of cases: sorting. Mentally or lexicographically, having the most important piece of information first will allow the correct item be found the fastest, or allow it to be discounted/ignored the quickest.
If you are looking for Bar, it is highly likely that you are already looking specifically for a particular functionality - say, the action - for Bar. As such, it is irrelevant which method you use, both will get you to the function you need.
Conversely, while it is likely you will want to look up all items that implement a particular functionality, it is much less likely you are going to ever need a complete listing of all functionality that an item employs; you will be targeting only one functionality for that item and will have that one functionality as the primary and concrete focus. Ergo, functionality comes first, followed by what item has that functionality.
We probably have slightly different work processes.
I'm more likely to be making "foo" functionally complete and then making "bar" complete than I am to be making all my dialogs functional then all my tabs/whatever.
This comes from TDD where I'm making a test pass for "foo", once done, I'll do the same for "bar".
Though it's even more likely these are different files entirely, rendering the arguments moot.
But also, sorting big endian automatically groups elements associated with common functions making search, completions, and snippets easier (if you use them). I'm torn
I was going to write something like this. You actually wrote about semantic order, but syntactically it is as much important e.g. it is easier to sort dates such as 2024-05-27 than 27.05.2024 in chronological order.