repeatedly omitting some tasks in a list of tasks

What version of Codex is running?

0.4.56

What subscription do you have?

plus

Which model were you using?

gpt-5.1-codex-max

What platform is your computer?

Linux 6.18.0-2-default x86_64 x86_64

What issue are you seeing?

Maybe there is an issue of codex not recognizing multiple tasks or paragraphs in a prompt?
When entering a prompt in the VS code extension, I often need to address several different items without submitting the prompt prematurely.
I sometimes try to make paragraphs using shift-return, but that collapses to a simple single newline in the chat log, so it doesn't seem to make a logical break (e.g. when submitting several logs for debugging, codex should be aware when one log ends and another starts). I try to use dashes or a) b) lists, but still my impression is that sometimes things get lost.

In a sense of prokastination, an ordered list of 5 topics may result in 2 or 3 being worked on, the other ones being ignored. I can repeat the open items in followup requests and the same topics get ignored again and again. In the extreme case, after a couple of such events, codex may even answer "we can keep task xy for later". It's a bit like asking your child "please do the dish washing" :-D

Suggestion:
keep multiple blank lines (Shift-Return) visible as empty lines in the chat log
Question:
Am I supposed to structure the prompt differently to distinguish multiple topics?

What steps can reproduce the bug?

In a complex task, let it ask for multiple feedback items and provide feedback on these topics in a single prompt. On top, while debugging one problem add unrelated tasks to a list of tasks. Separate different tasks with dashes or enumerate them (makes no difference as far as I can tell). I don't have a good test case to share

What is the expected behavior?

it should address every information given in the prompt, not just the "hottest" topics.

Additional information

context window usually stays <70% as indicated in vs code