Eli Zaretskii writes:
 > A reproducible test case, preferably starting with "emacs -Q", would
 > be greatly appreciated.
 > 

Note that both the crash and the bug (mentioned further below) were
not reproducible on a virtual Windows XP.

To reproduce this crash:

1. Open attached test case with: emacs -Q --eval "(setq-default bidi-display-reordering t)" wrap_crash.org
   It should open in org-mode by default.

2. Emacs _sometimes_ would render it like this:

* First...
                           ...Second *

 If not, repeat step one until it does, or use something similar to
 the command I provided in my last message, etc.

3. Once you see the above, do: M-x visual-line-mode [ENTER] C-n C-n

 > > It seems that Emacs _sometimes_ renders the above example in org-mode
 > > as,
 > > 
 > > * First...
 > > * Second
 > > 
 > > While in other invocations the same file is rendered as:
 > > 
 > > * First...
 > >                              Second *
 > 
 > Sorry, I don't understand.  Your original example included a line in
 > ARABIC and an empty line, in addition to "First" and "second" lines.
 > Are we talking about a different example now? 

No, it's the same example just viewed using org-mode (which considers
lines that start with `*' as headers, and does some folding).

 > For the example you show now:
 > 
 >  * First
 >  * Second
 > 
 > the correct rendering is what you'd expect:
 > 
 >  * First
 >  * Second

Would this apply to org-mode wrt to the attached testcase?

As a side bug, if you position cursor on "* First" in the testcase
then press [TAB] you will see that "* Second" gets deformed to:

d *...

Testcase and backtrace attached. Thanks.