diff --git a/diffrax/_root_finder/_verychord.py b/diffrax/_root_finder/_verychord.py
index 921a6dd4..385a2cb9 100644
--- a/diffrax/_root_finder/_verychord.py
+++ b/diffrax/_root_finder/_verychord.py
@@ -162,7 +162,7 @@ def terminate(
         converged = _converged(factor, self.kappa)
         terminate = at_least_two & (small | diverged | converged)
         terminate_result = optx.RESULTS.where(
-            jnp.invert(small) & (diverged | jnp.invert(converged)),
+            at_least_two & jnp.invert(small) & (diverged | jnp.invert(converged)),
             optx.RESULTS.nonlinear_divergence,
             optx.RESULTS.successful,
         )
diff --git a/pyproject.toml b/pyproject.toml
index 880c50c9..01b2e8f8 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -19,7 +19,7 @@ classifiers = [
   "Topic :: Scientific/Engineering :: Information Analysis",
   "Topic :: Scientific/Engineering :: Mathematics"
 ]
-dependencies = ["jax>=0.4.38", "jaxtyping>=0.2.24", "typing_extensions>=4.5.0", "equinox>=0.11.10", "lineax>=0.0.5", "optimistix>=0.0.10", "wadler_lindig>=0.1.1"]
+dependencies = ["jax>=0.4.38", "jaxtyping>=0.2.24", "typing_extensions>=4.5.0", "equinox>=0.11.10", "lineax>=0.0.5", "optimistix>=0.1.0", "wadler_lindig>=0.1.1"]
 description = "GPU+autodiff-capable ODE/SDE/CDE solvers written in JAX."
 keywords = ["jax", "dynamical-systems", "differential-equations", "deep-learning", "equinox", "neural-differential-equations", "diffrax"]
 license = {file = "LICENSE"}
diff --git a/test/test_very_chord.py b/test/test_very_chord.py
index c0d0287a..17424630 100644
--- a/test/test_very_chord.py
+++ b/test/test_very_chord.py
@@ -24,6 +24,9 @@ def _fn2(x, args):
 @jax.jit
 def _fn3(x, args):
     mlp = eqx.nn.MLP(4, 4, 256, 2, key=jr.PRNGKey(678))
+    dynamic, static = eqx.partition(mlp, eqx.is_array)
+    dynamic = jtu.tree_map(lambda x: x * 0.1, dynamic)
+    mlp = eqx.combine(dynamic, static)
     return mlp(x) - x