diff --git a/doc/changelog.rst b/doc/changelog.rst
index 25ac5a604..0f4ce46b0 100644
--- a/doc/changelog.rst
+++ b/doc/changelog.rst
@@ -5,6 +5,19 @@
 Changelog
 =========
 
+2.8.4 (October 11, 2021)
+------------------------
+Fixes
+~~~~~
+- Pickling of ``util.Matrix`` resulted in incomplete state after unpickling, which would
+  cause an exception when you tried to print the matrix
+
+Documentation
+~~~~~~~~~~~~~
+- New tutorial on fitting PDFs that depend on a conditional variable
+- Fixed JAX tutorial, adapting to change in their interface
+- Extended documentation of cost functions
+
 2.8.3 (September 3, 2021)
 -------------------------
 
diff --git a/src/iminuit/cost.py b/src/iminuit/cost.py
index 5db351cc2..3845d3fda 100644
--- a/src/iminuit/cost.py
+++ b/src/iminuit/cost.py
@@ -1,21 +1,25 @@
 """
 Standard cost functions to minimize for statistical fits.
 
+We provide these for convenience, so that you do not have to write your own for standard fits.
+The cost functions optionally use Numba to accelerate some calculations, if Numba is
+installed.
+
 What to use when
 ----------------
 - Fit a normalised probability density to data
   - Data is not binned: UnbinnedNLL
-  - Data is binned: BinnedNLL; also supports histogram of weighted samples
+  - Data is binned: BinnedNLL, also supports histogram of weighted samples
 - Fit a density to data, density is not normalised
   - Data is not binned: ExtendedUnbinnedNLL
-  - Data is binned: ExtendedBinnedNLL; also supports histogram of weighted samples
+  - Data is binned: ExtendedBinnedNLL, also supports histogram of weighted samples
 - Fit of a function f(x) to (x, y, yerror) pairs with normal-distributed fluctuations
   (x can be multi-dimensional)
   - y values contain no outliers: LeastSquares
   - y values contain outliers: LeastSquares with "soft_l1" loss function
 
-Cost functions are addable
---------------------------
+Combining cost functions
+------------------------
 All cost functions can be added, which generates a new combined cost function.
 Parameters with the same name are shared between component cost functions. Use this to
 constrain one or several parameters with different data sets and using different
@@ -24,10 +28,9 @@
 
 Notes
 -----
-The cost functions defined here should be preferred over custom implementations. They
-have been optimized with knowledge about implementation details of Minuit to give the
-highest accucary and the most robust results. They are partially accelerated with numba,
-if numba is available.
+The cost functions defined here have been optimized with knowledge about implementation
+details of Minuit to give the highest accucary and the most robust results, so they
+should perform well. If you have trouble with your own implementations, try these.
 
 The binned versions of the log-likelihood fits support weighted samples. For each bin
 of the histogram, the sum of weights and the sum of squared weights is needed then, see
@@ -852,9 +855,6 @@ def _call(self, args):
         return np.einsum("i,ij,j", delta, self._covinv, delta)
 
 
-GaussianPenalty = NormalConstraint
-
-
 def _norm(value):
     value = np.atleast_1d(value)
     dtype = value.dtype
diff --git a/src/iminuit/util.py b/src/iminuit/util.py
index cec30f37c..d2fa7d8b7 100644
--- a/src/iminuit/util.py
+++ b/src/iminuit/util.py
@@ -1,4 +1,9 @@
-"""Data classes and utilities used by :class:`iminuit.Minuit`."""
+"""
+Data classes and utilities used by :class:`iminuit.Minuit`.
+
+You can look up the interface of data classes that iminuit uses here. Also of interest
+for users are :func:`propagate` and :func:`make_with_signature`.
+"""
 import inspect
 from collections import OrderedDict
 from argparse import Namespace
diff --git a/src/iminuit/version.py b/src/iminuit/version.py
index f3ef59ea9..04b9ca42b 100644
--- a/src/iminuit/version.py
+++ b/src/iminuit/version.py
@@ -7,7 +7,7 @@
 # - Increase MAINTENANCE when fixing bugs without adding features
 # - During development, add suffix .devN with N >= 0
 # - For release candidates, add suffix .rcN with N >= 0
-version = "2.8.3"
+version = "2.8.4"
 
 # We list the corresponding ROOT version of the C++ Minuit2 library here
 root_version = "v6-25-01-1694-g187368db19"
diff --git a/tests/test_issue.py b/tests/test_issue.py
index c540ebaf6..d76de72b5 100644
--- a/tests/test_issue.py
+++ b/tests/test_issue.py
@@ -109,7 +109,10 @@ def test_issue_687():
     m = Minuit(fcn, start)
 
     m.migrad()
+    s_m = str(m)
 
     s = pickle.dumps(m)
     m2 = pickle.loads(s)
-    print(m2)  # this used to fail
+
+    s_m2 = str(m2)  # this used to fail
+    assert s_m == s_m2