How to create numbered examples potentially with letter suffixes, e.g. using enumitem?

Question

I have a weird numbering problem in a manuscript that uses sequentially numbered text examples. When an example demonstrates something which is e.g. wrong and later to be corrected they have suffixes a, b, and so on. In between it happens that there is text and possibly other examples which continues the original numbering, e.g:

(1) ... example ...

... text ...

(2a) ... bad example ...

... enlightening text and example:

(3) ... example ...

... after which we might return to the earlier bad example (2a) and show the correction

(2b) ... good example ...

... back to normal with more examples:

(4) ... example ...

I was asked how one would do this in a smart way in LaTeX. I have tried using the enumitem package as follows, but ran into trouble:

\documentclass{article}

\usepackage{enumitem}

\newlist{examples}{enumerate}{1}
\setlist[examples]{label=(\arabic*)}
\newlist{subexamples}{enumerate}{1}

\begin{document}
First an example, which we call 1.
\begin{examples}
  \item example
\end{examples}
Then a bad example which we would like to call 2a: 
\addtocounter{examplesi}{1}
\begin{subexamples}[series=aname, label=(\theexamplesi\alph*)]
  \item example
\end{subexamples}
We learn something new in examples 3 and 4 (with wrong labels):
\begin{examples}[resume]
  \item example
  \item example
\end{examples}
And we get back to the bad example, which should be 2b: 
\begin{subexamples}[resume*=aname]
  \item example
\end{subexamples}

\end{document}

The above example produces the labels 1, 2a, 2, 3, 3b. The goal was 1, 2a, 3, 4, 2b.

One should be able to reference the individual items, and it should be flexible enough to have more than a single "hanging" example...

Answer 1

Well, I found one way to do it; I wouldn't call it satisfactory but it works. The cons are: you can't mix examples with suffixes and examples without in a single list, and the siffixes a, b, ... are 'hard coded' whereas the main counter is 'automatic'. Also, I would think it could be prettier as the referencing is slightly messy. Here goes:

\documentclass{article}

\usepackage{enumitem}
\newlist{examples}{enumerate}{1}
\setlist[examples]{label=(\arabic*), ref=\arabic*, resume}

\begin{document}

\section{On numbered examples}
First, an example:
\begin{examples}
   \item\label{first} \dots example \dots
\end{examples}
This can be referenced as expected: check out (\ref{first}).
Now, a second example which we can elaborate on later is
\begin{examples}[label=(\arabic*a), ref=\arabic*]
   \item\label{second} \dots example \dots
\end{examples}
The suffix in (\ref{second}a) hard coded as \verb+label=(\arabic*a)+
in the options for this one example environment.

We learn something new in examples 3 and 4, and use the
environment as is:
\begin{examples}
   \item example
   \item example
\end{examples}
And we get back to previous example in examples 2b and 2c:
\begin{examples}
  \item[(\ref{second}b)] \dots example \dots
  \item[(\ref{second}c)] \dots example \dots
\end{examples}
Since \verb+\item+ is used with \verb+[]+ the counter is not
incremented. These are referenced as e.g. \verb+(\ref{second}b)+.
And we test that the counter is as it should be:
\begin{examples}
   \item \dots example \dots
\end{examples}

\end{document}