diff --git a/.github/workflows/tests.yml b/.github/workflows/tests.yml
index c4e7fc9..8d5cae2 100644
--- a/.github/workflows/tests.yml
+++ b/.github/workflows/tests.yml
@@ -44,5 +44,18 @@ jobs:
         with:
           python-version: '3.11'
           cache: 'pip'
-      - run: pip install -r requirements.txt
-      - run: python -m pytest tests
\ No newline at end of file
+
+      - name: Install deps
+        run: pip install -r requirements.txt
+
+      - name: Tests with numba
+        run: coverage run --source=tsbrowse -m pytest -x tests
+    
+      - name: Tests with coverage (no numba)
+        run: TSBROWSE_DISABLE_NUMBA=1 coverage run --source=tsbrowse -m pytest -x tests
+
+      - name: Upload coverage
+        env:
+          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}
+        run: |
+          coveralls --service=github
\ No newline at end of file
diff --git a/requirements.txt b/requirements.txt
index 905dce6..ebc1592 100644
--- a/requirements.txt
+++ b/requirements.txt
@@ -1,9 +1,8 @@
 appdirs
 click
+coverage
+coveralls
 daiquiri
-# It's not clear why we need Dask here, but tests are failing
-# otherwise. Pinning to arbitrary working version
-dask[dataframe]==2024.4.1
 datashader
 diskcache
 hvplot
diff --git a/tsbrowse/jit.py b/tsbrowse/jit.py
new file mode 100644
index 0000000..7ec9c6e
--- /dev/null
+++ b/tsbrowse/jit.py
@@ -0,0 +1,56 @@
+import functools
+import logging
+import os
+
+import numba
+
+logger = logging.getLogger(__name__)
+
+_DISABLE_NUMBA = os.environ.get("TSBROWSE_DISABLE_NUMBA", "0")
+
+try:
+    ENABLE_NUMBA = {"0": True, "1": False}[_DISABLE_NUMBA]
+except KeyError as e:  # pragma: no cover
+    raise KeyError(
+        "Environment variable 'TSBROWSE_DISABLE_NUMBA' must be '0' or '1'"
+    ) from e
+
+# We will mostly be using disable numba for debugging and running tests for
+# coverage, so raise a loud warning in case this is being used accidentally.
+
+if not ENABLE_NUMBA:
+    logger.warning(
+        "numba globally disabled for tsbrowse; performance will be drastically"
+        " reduced."
+    )
+
+
+DEFAULT_NUMBA_ARGS = {
+    "nopython": True,
+    "cache": True,
+}
+
+
+def numba_njit(**numba_kwargs):
+    def _numba_njit(func):
+        @functools.wraps(func)
+        def wrapper(*args, **kwargs):
+            return func(*args, **kwargs)  # pragma: no cover
+
+        if ENABLE_NUMBA:  # pragma: no cover
+            combined_kwargs = {**DEFAULT_NUMBA_ARGS, **numba_kwargs}
+            return numba.jit(**combined_kwargs)(func)
+        else:
+            return func
+
+    return _numba_njit
+
+
+def numba_jitclass(spec):
+    def _numba_jitclass(cls):
+        if ENABLE_NUMBA:  # pragma: no cover
+            return numba.experimental.jitclass(spec)(cls)
+        else:
+            return cls
+
+    return _numba_jitclass
diff --git a/tsbrowse/model.py b/tsbrowse/model.py
index 283c9e3..eb273a4 100644
--- a/tsbrowse/model.py
+++ b/tsbrowse/model.py
@@ -8,9 +8,9 @@
 import pandas as pd
 import tskit
 
+from . import jit
 from .cache import disk_cache
 
-
 logger = daiquiri.getLogger("tsbrowse")
 
 spec = [
@@ -28,7 +28,7 @@
 ]
 
 
-@numba.experimental.jitclass(spec)
+@jit.numba_jitclass(spec)
 class TreePosition:
     def __init__(
         self,
@@ -91,7 +91,7 @@ def alloc_tree_position(ts):
     )
 
 
-@numba.njit
+@jit.numba_njit()
 def _compute_per_tree_stats(
     tree_pos, num_trees, num_nodes, nodes_time, edges_parent, edges_child
 ):
@@ -163,7 +163,7 @@ def compute_per_tree_stats(ts):
     )
 
 
-@numba.njit
+@jit.numba_njit()
 def _compute_mutation_parent_counts(mutations_parent):
     N = mutations_parent.shape[0]
     num_parents = np.zeros(N, dtype=np.int32)
@@ -176,7 +176,7 @@ def _compute_mutation_parent_counts(mutations_parent):
     return num_parents
 
 
-@numba.njit
+@jit.numba_njit()
 def _compute_mutation_inheritance_counts(
     tree_pos,
     num_nodes,
@@ -259,7 +259,7 @@ def compute_mutation_counts(ts):
     return MutationCounts(num_parents, num_inheritors, num_descendants)
 
 
-@numba.njit
+@jit.numba_njit()
 def _compute_population_mutation_counts(
     tree_pos,
     num_nodes,
@@ -398,7 +398,7 @@ def _repr_html_(self):
         return self.summary_df._repr_html_()
 
     @staticmethod
-    @numba.njit
+    @jit.numba_njit()
     def child_bounds(num_nodes, edges_left, edges_right, edges_child):
         num_edges = edges_left.shape[0]
         child_left = np.zeros(num_nodes, dtype=np.float64) + np.inf