I had an issue in Rmarkdown which I ultimately tracked down to Pandoc.

To reproduce

1. Construct the following Markdown file, e.g., testfile.md:

   Text before equations.
   \begin{subequations}
   \begin{align}
   a\\
   b
   \end{align}
   \end{subequations}
   Text after equations.
   
   Text before equations.
   \begin{align}
   a\\
   b
   \end{align}
   Text after equations.

2. Run pandoc testfile.md -o testfile.tex.

Expected result

The following LaTeX file:

Text before equations. \begin{subequations}
\begin{align}
a\\
b
\end{align}
\end{subequations} Text after equations.

Text before equations. \begin{align}
a\\
b
\end{align} Text after equations.

Actual result

Text before equations.

\begin{subequations}
\begin{align}
a\\
b
\end{align}
\end{subequations}

Text after equations.

Text before equations. \begin{align}
a\\
b
\end{align} Text after equations.

Thus, spurious paragraph breaks are created before \begin{subequations} and after \end{subequations} but not so with the align environment.

Workaround

For now, this can be worked around by replacing \begin{subequations} and \end{subequations} with LaTeX macros:

\def\bse{\begin{subequations}}
\def\ese{\end{subequations}}

Notes

From keyword searching through the source code, I suspect that at least a part of the fix would be to add "subequations" to the list below. However, I know neither Haskell nor Pandoc's internals, so I don't know what else is involved.