diff --git a/doc/getting-started-guide/faq.rst b/doc/getting-started-guide/faq.rst
index 58d5448cdf5..49d7f5e4873 100644
--- a/doc/getting-started-guide/faq.rst
+++ b/doc/getting-started-guide/faq.rst
@@ -146,6 +146,9 @@ for conflicts between ``attrs`` when combining arrays and datasets, unless
 explicitly requested with the option ``compat='identical'``. The guiding
 principle is that metadata should not be allowed to get in the way.
 
+In general xarray uses the capabilities of the backends for reading and writing
+attributes. That has some implications on roundtripping. One example for such inconsistency is that size-1 lists will roundtrip as single element (for netcdf4 backends).
+
 What other netCDF related Python libraries should I know about?
 ---------------------------------------------------------------
 
diff --git a/doc/user-guide/data-structures.rst b/doc/user-guide/data-structures.rst
index e5e89b0fbbd..9a2f26ec7b5 100644
--- a/doc/user-guide/data-structures.rst
+++ b/doc/user-guide/data-structures.rst
@@ -40,7 +40,8 @@ alignment, building on the functionality of the ``index`` found on a pandas
 DataArray objects also can have a ``name`` and can hold arbitrary metadata in
 the form of their ``attrs`` property. Names and attributes are strictly for
 users and user-written code: xarray makes no attempt to interpret them, and
-propagates them only in unambiguous cases
+propagates them only in unambiguous cases. For reading and writing attributes
+xarray relies on the capabilities of the supported backends.
 (see FAQ, :ref:`approach to metadata`).
 
 .. _creating a dataarray:
diff --git a/doc/whats-new.rst b/doc/whats-new.rst
index 659063f2cbf..baa57e7bb27 100644
--- a/doc/whats-new.rst
+++ b/doc/whats-new.rst
@@ -43,6 +43,9 @@ Bug fixes
 Documentation
 ~~~~~~~~~~~~~
 
+- Mention attribute peculiarities in docs/docstrings (:issue:`4798`, :pull:`9700`).
+  By `Kai Mühlbauer <https://github.com/kmuehlbauer>`_.
+
 
 Internal Changes
 ~~~~~~~~~~~~~~~~
diff --git a/xarray/core/dataarray.py b/xarray/core/dataarray.py
index c6bc082f5ed..f04cdac3a74 100644
--- a/xarray/core/dataarray.py
+++ b/xarray/core/dataarray.py
@@ -347,6 +347,7 @@ class DataArray(
     attrs : dict_like or None, optional
         Attributes to assign to the new instance. By default, an empty
         attribute dictionary is initialized.
+        (see FAQ, :ref:`approach to metadata`)
     indexes : py:class:`~xarray.Indexes` or dict-like, optional
         For internal use only. For passing indexes objects to the
         new DataArray, use the ``coords`` argument instead with a
diff --git a/xarray/core/dataset.py b/xarray/core/dataset.py
index bc9360a809d..3cbeed3f7f4 100644
--- a/xarray/core/dataset.py
+++ b/xarray/core/dataset.py
@@ -596,6 +596,7 @@ class Dataset(
 
     attrs : dict-like, optional
         Global attributes to save on this dataset.
+        (see FAQ, :ref:`approach to metadata`)
 
     Examples
     --------
diff --git a/xarray/core/variable.py b/xarray/core/variable.py
index 13053faff58..16e292796a7 100644
--- a/xarray/core/variable.py
+++ b/xarray/core/variable.py
@@ -401,6 +401,7 @@ def __init__(
         attrs : dict_like or None, optional
             Attributes to assign to the new variable. If None (default), an
             empty attribute dictionary is initialized.
+            (see FAQ, :ref:`approach to metadata`)
         encoding : dict_like or None, optional
             Dictionary specifying how to encode this array's data into a
             serialized format like netCDF4. Currently used keys (for netCDF)