> Do you have a test case to reproduce this bug?

No.

See the original report.  It points you to my
code that fixes the problem by sprinkling
`constrain-to-field' everywhere.

To give an idea, attached are two functions,
my version and the vanilla version.  Ediff
them.  See the additions of this:

 (constrain-to-field nil orig)

I believe that the vanilla function needs
similar protection.  For you to decide.

That's just one thingatpt.el function.  I
also added `constraint-to-field to other
such functions.

See thingatpt+.el if you want to see the
occurrences, to maybe save some time when
checking where thingatpt.el might need the
same treatment.

I think you get the idea.  Thing-at-point
should respect fields.  Do it or don't do
it; your choice.