diff --git a/pixi.lock b/pixi.lock
index 6b3a38bc..167028d6 100644
--- a/pixi.lock
+++ b/pixi.lock
@@ -190,7 +190,7 @@ environments:
       - conda: https://conda.anaconda.org/conda-forge/noarch/websocket-client-1.9.0-pyhd8ed1ab_0.conda
       - conda: https://conda.anaconda.org/conda-forge/noarch/zipp-3.23.1-pyhcf101f3_0.conda
       - pypi: .
-      - pypi: git+https://github.com/easyscience/corelib?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
+      - pypi: git+https://github.com/easyscience/corelib.git?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
       - pypi: https://files.pythonhosted.org/packages/00/bb/90ba423612b6aa0adccc6b1874bcd4a9b44b660c0c16f346611e00f64ac3/backrefs-7.0-py313-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/01/7c/fa07d3da2b6253eb8474be16eab2eadf670460e364ccc895ca7ff388ee30/oscrypto-1.3.0-py2.py3-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/04/11/432f32f8097b03e3cd5fe57e88efb685d964e2e5178a48ed61e841f7fdce/pyyaml_env_tag-1.1-py3-none-any.whl
@@ -501,7 +501,7 @@ environments:
       - conda: https://conda.anaconda.org/conda-forge/osx-arm64/zeromq-4.3.5-h4818236_10.conda
       - conda: https://conda.anaconda.org/conda-forge/osx-arm64/zstd-1.5.7-hbf9d68e_6.conda
       - pypi: .
-      - pypi: git+https://github.com/easyscience/corelib?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
+      - pypi: git+https://github.com/easyscience/corelib.git?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
       - pypi: https://files.pythonhosted.org/packages/00/bb/90ba423612b6aa0adccc6b1874bcd4a9b44b660c0c16f346611e00f64ac3/backrefs-7.0-py313-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/01/7c/fa07d3da2b6253eb8474be16eab2eadf670460e364ccc895ca7ff388ee30/oscrypto-1.3.0-py2.py3-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/04/11/432f32f8097b03e3cd5fe57e88efb685d964e2e5178a48ed61e841f7fdce/pyyaml_env_tag-1.1-py3-none-any.whl
@@ -802,7 +802,7 @@ environments:
       - conda: https://conda.anaconda.org/conda-forge/win-64/zeromq-4.3.5-h507cc87_10.conda
       - conda: https://conda.anaconda.org/conda-forge/win-64/zstd-1.5.7-h534d264_6.conda
       - pypi: .
-      - pypi: git+https://github.com/easyscience/corelib?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
+      - pypi: git+https://github.com/easyscience/corelib.git?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
       - pypi: https://files.pythonhosted.org/packages/00/bb/90ba423612b6aa0adccc6b1874bcd4a9b44b660c0c16f346611e00f64ac3/backrefs-7.0-py313-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/01/7c/fa07d3da2b6253eb8474be16eab2eadf670460e364ccc895ca7ff388ee30/oscrypto-1.3.0-py2.py3-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/04/11/432f32f8097b03e3cd5fe57e88efb685d964e2e5178a48ed61e841f7fdce/pyyaml_env_tag-1.1-py3-none-any.whl
@@ -1125,7 +1125,7 @@ environments:
       - conda: https://conda.anaconda.org/conda-forge/noarch/websocket-client-1.9.0-pyhd8ed1ab_0.conda
       - conda: https://conda.anaconda.org/conda-forge/noarch/zipp-3.23.1-pyhcf101f3_0.conda
       - pypi: .
-      - pypi: git+https://github.com/easyscience/corelib?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
+      - pypi: git+https://github.com/easyscience/corelib.git?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
       - pypi: https://files.pythonhosted.org/packages/01/7c/fa07d3da2b6253eb8474be16eab2eadf670460e364ccc895ca7ff388ee30/oscrypto-1.3.0-py2.py3-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/04/11/432f32f8097b03e3cd5fe57e88efb685d964e2e5178a48ed61e841f7fdce/pyyaml_env_tag-1.1-py3-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/06/8a/5e156e31ba656ce93c1cc895dd8f051ec351cb382940dca655aaec475005/python_bidi-0.6.9-cp311-cp311-manylinux_2_17_x86_64.manylinux2014_x86_64.whl
@@ -1430,7 +1430,7 @@ environments:
       - conda: https://conda.anaconda.org/conda-forge/osx-arm64/zeromq-4.3.5-h4818236_10.conda
       - conda: https://conda.anaconda.org/conda-forge/osx-arm64/zstd-1.5.7-hbf9d68e_6.conda
       - pypi: .
-      - pypi: git+https://github.com/easyscience/corelib?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
+      - pypi: git+https://github.com/easyscience/corelib.git?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
       - pypi: https://files.pythonhosted.org/packages/01/7c/fa07d3da2b6253eb8474be16eab2eadf670460e364ccc895ca7ff388ee30/oscrypto-1.3.0-py2.py3-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/04/11/432f32f8097b03e3cd5fe57e88efb685d964e2e5178a48ed61e841f7fdce/pyyaml_env_tag-1.1-py3-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/07/c7/deb8c5e604404dbf10a3808a858946ca3547692ff6316b698945bb72177e/python_socketio-5.16.1-py3-none-any.whl
@@ -1725,7 +1725,7 @@ environments:
       - conda: https://conda.anaconda.org/conda-forge/win-64/zeromq-4.3.5-h507cc87_10.conda
       - conda: https://conda.anaconda.org/conda-forge/win-64/zstd-1.5.7-h534d264_6.conda
       - pypi: .
-      - pypi: git+https://github.com/easyscience/corelib?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
+      - pypi: git+https://github.com/easyscience/corelib.git?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
       - pypi: https://files.pythonhosted.org/packages/01/7c/fa07d3da2b6253eb8474be16eab2eadf670460e364ccc895ca7ff388ee30/oscrypto-1.3.0-py2.py3-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/02/05/d60c732b56da5085175c07c74b2df4e6d181b0c9a61e1691474f06ef4b39/lxml-6.1.0-cp311-cp311-win_amd64.whl
       - pypi: https://files.pythonhosted.org/packages/04/11/432f32f8097b03e3cd5fe57e88efb685d964e2e5178a48ed61e841f7fdce/pyyaml_env_tag-1.1-py3-none-any.whl
@@ -2048,7 +2048,7 @@ environments:
       - conda: https://conda.anaconda.org/conda-forge/noarch/websocket-client-1.9.0-pyhd8ed1ab_0.conda
       - conda: https://conda.anaconda.org/conda-forge/noarch/zipp-3.23.1-pyhcf101f3_0.conda
       - pypi: .
-      - pypi: git+https://github.com/easyscience/corelib?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
+      - pypi: git+https://github.com/easyscience/corelib.git?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
       - pypi: https://files.pythonhosted.org/packages/00/bb/90ba423612b6aa0adccc6b1874bcd4a9b44b660c0c16f346611e00f64ac3/backrefs-7.0-py313-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/01/7c/fa07d3da2b6253eb8474be16eab2eadf670460e364ccc895ca7ff388ee30/oscrypto-1.3.0-py2.py3-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/04/11/432f32f8097b03e3cd5fe57e88efb685d964e2e5178a48ed61e841f7fdce/pyyaml_env_tag-1.1-py3-none-any.whl
@@ -2359,7 +2359,7 @@ environments:
       - conda: https://conda.anaconda.org/conda-forge/osx-arm64/zeromq-4.3.5-h4818236_10.conda
       - conda: https://conda.anaconda.org/conda-forge/osx-arm64/zstd-1.5.7-hbf9d68e_6.conda
       - pypi: .
-      - pypi: git+https://github.com/easyscience/corelib?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
+      - pypi: git+https://github.com/easyscience/corelib.git?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
       - pypi: https://files.pythonhosted.org/packages/00/bb/90ba423612b6aa0adccc6b1874bcd4a9b44b660c0c16f346611e00f64ac3/backrefs-7.0-py313-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/01/7c/fa07d3da2b6253eb8474be16eab2eadf670460e364ccc895ca7ff388ee30/oscrypto-1.3.0-py2.py3-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/04/11/432f32f8097b03e3cd5fe57e88efb685d964e2e5178a48ed61e841f7fdce/pyyaml_env_tag-1.1-py3-none-any.whl
@@ -2660,7 +2660,7 @@ environments:
       - conda: https://conda.anaconda.org/conda-forge/win-64/zeromq-4.3.5-h507cc87_10.conda
       - conda: https://conda.anaconda.org/conda-forge/win-64/zstd-1.5.7-h534d264_6.conda
       - pypi: .
-      - pypi: git+https://github.com/easyscience/corelib?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
+      - pypi: git+https://github.com/easyscience/corelib.git?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
       - pypi: https://files.pythonhosted.org/packages/00/bb/90ba423612b6aa0adccc6b1874bcd4a9b44b660c0c16f346611e00f64ac3/backrefs-7.0-py313-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/01/7c/fa07d3da2b6253eb8474be16eab2eadf670460e364ccc895ca7ff388ee30/oscrypto-1.3.0-py2.py3-none-any.whl
       - pypi: https://files.pythonhosted.org/packages/04/11/432f32f8097b03e3cd5fe57e88efb685d964e2e5178a48ed61e841f7fdce/pyyaml_env_tag-1.1-py3-none-any.whl
@@ -8301,8 +8301,9 @@ packages:
   name: easyreflectometry
   requires_dist:
   - bumps
-  - easyscience @ git+https://github.com/easyscience/corelib@develop
+  - easyscience @ git+https://github.com/easyscience/corelib.git@develop
   - orsopy
+  - plotly
   - pooch
   - refl1d>=1.0.0
   - refnx
@@ -8344,7 +8345,7 @@ packages:
   - validate-pyproject[all] ; extra == 'dev'
   - versioningit ; extra == 'dev'
   requires_python: '>=3.11'
-- pypi: git+https://github.com/easyscience/corelib?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
+- pypi: git+https://github.com/easyscience/corelib.git?rev=develop#aadbd4891b94f6aa18187d48be8c2ab6f81113b0
   name: easyscience
   version: 2.3.1+dev8
   requires_dist:
diff --git a/pyproject.toml b/pyproject.toml
index ac183b21..596b7f27 100644
--- a/pyproject.toml
+++ b/pyproject.toml
@@ -23,7 +23,8 @@ classifiers = [
 ]
 requires-python = '>=3.11'
 dependencies = [
-  'easyscience @ git+https://github.com/easyscience/corelib.git@develop',
+  # 'easyscience @ git+https://github.com/easyscience/corelib.git@develop',
+  'easyscience @ git+https://github.com/easyscience/corelib.git@bayesian_extend',
   # 'easyscience',
   'scipp',
   'refnx',
diff --git a/src/easyreflectometry/analysis/bayesian.py b/src/easyreflectometry/analysis/bayesian.py
index 39389a84..17ab252c 100644
--- a/src/easyreflectometry/analysis/bayesian.py
+++ b/src/easyreflectometry/analysis/bayesian.py
@@ -1332,8 +1332,9 @@ def load_posterior(path: str, skip: int = 0) -> 'PosteriorResults':
     """Reload a trace saved by :func:`save_posterior` into a
     :class:`PosteriorResults`.
 
-    The returned object's ``sampler_state`` can be fed back into
-    ``MultiFitter.mcmc_sample(..., resume_state=...)`` to extend the chain.
+    The returned object's ``sampler_state`` can be fed back into the core
+    ``Sampler`` (via ``Sampler.load_state(...)`` / ``Sampler.extend(...)``)
+    to extend the chain.
 
     :param path: File path prefix used in :func:`save_posterior`.
     :type path: str
diff --git a/src/easyreflectometry/fitting.py b/src/easyreflectometry/fitting.py
index 83a3f6eb..f90211ff 100644
--- a/src/easyreflectometry/fitting.py
+++ b/src/easyreflectometry/fitting.py
@@ -10,6 +10,7 @@
 import scipp as sc
 from easyscience.fitting import AvailableMinimizers
 from easyscience.fitting import FitResults
+from easyscience.fitting import Sampler
 from easyscience.fitting.multi_fitter import MultiFitter as EasyScienceMultiFitter
 
 from easyreflectometry.data import DataSet1D
@@ -419,14 +420,22 @@ def mcmc_sample(
             y.append(y_eff)
             dy.append(weights)
 
-        # Delegate the actual BUMPS/DREAM sampling to the core MultiFitter
+        # Delegate the actual BUMPS/DREAM sampling to the core ``Sampler``.
+        # The core API moved from ``MultiFitter.mcmc_sample()`` to a dedicated
+        # ``Sampler`` class: construct it with the configured fitter and the
+        # bound data, then call ``sample()``. ``Sampler`` handles the
+        # multi-dataset reshaping internally.
         sampler_kwargs = {}
         if initializer is not None:
             sampler_kwargs['init'] = initializer
-        return self.easy_science_multi_fitter.mcmc_sample(
+
+        sampler = Sampler(
+            self.easy_science_multi_fitter,
             x=x,
             y=y,
             weights=dy,
+        )
+        results = sampler.sample(
             samples=samples,
             burn=burn,
             thin=thin,
@@ -435,6 +444,12 @@ def mcmc_sample(
             progress_callback=progress_callback,
             abort_test=abort_test,
         )
+        return {
+            'draws': results.draws,
+            'param_names': results.param_names,
+            'state': results.state,
+            'logp': results.logp,
+        }
 
     @property
     def chi2(self) -> float | None:
diff --git a/tests/test_fitting.py b/tests/test_fitting.py
index e4f938c8..2c519867 100644
--- a/tests/test_fitting.py
+++ b/tests/test_fitting.py
@@ -4,6 +4,7 @@
 
 import os
 from unittest.mock import MagicMock
+from unittest.mock import patch
 
 import numpy as np
 import pytest
@@ -808,6 +809,44 @@ def _fake_fit(*, x, y, weights):
 # ---------------------------------------------------------------------------
 
 
+def _fake_sampling_results(draws=None, param_names=None, state=None, logp=None):
+    """Build a stand-in for the core ``SamplingResults`` returned by ``Sampler.sample``."""
+    res = MagicMock()
+    res.draws = np.ones((10, 2)) if draws is None else draws
+    res.param_names = ['a', 'b'] if param_names is None else param_names
+    res.state = state
+    res.logp = logp
+    return res
+
+
+def _patch_sampler(capture, results=None):
+    """Patch ``easyreflectometry.fitting.Sampler`` and capture its call args.
+
+    Records the constructor's ``(x, y, weights)`` and the ``sample()``
+    hyperparameters into the ``capture`` dict, and returns ``results`` (a
+    fake ``SamplingResults``) from ``sample()``.
+    """
+    results = results if results is not None else _fake_sampling_results()
+
+    def _ctor(fitter, *, x, y, weights, **kwargs):
+        capture['fitter'] = fitter
+        capture['x'] = x
+        capture['y'] = y
+        capture['weights'] = weights
+        capture.update(kwargs)  # e.g. sampler_kwargs if passed to the ctor
+        instance = MagicMock()
+
+        def _sample(**sample_kwargs):
+            capture.update(sample_kwargs)
+            return results
+
+        instance.sample = MagicMock(side_effect=_sample)
+        capture['instance'] = instance
+        return instance
+
+    return patch('easyreflectometry.fitting.Sampler', side_effect=_ctor)
+
+
 class TestMCMCSampleRequiresBumpsEngine:
     """mcmc_sample() must raise when the core engine is not a BUMPS instance."""
 
@@ -824,138 +863,129 @@ def test_raises_runtime_error_when_not_bumps(self):
         with pytest.raises(RuntimeError, match='Bayesian sampling requires a BUMPS minimizer'):
             fitter.mcmc_sample(data)
 
-    def test_wrapper_check_runs_before_core_mcmc_sample(self):
-        """The wrapper-level guard must fire before delegating to the core sampler.
+    def test_wrapper_check_runs_before_sampler(self):
+        """The wrapper-level guard must fire before constructing the core ``Sampler``.
 
-        Replace the core ``mcmc_sample`` with a sentinel that would record any call;
-        the guard should raise without invoking it.
+        Patch ``Sampler`` with a sentinel that would record any instantiation;
+        the guard should raise without ever building it.
         """
         model = Model()
         model.interface = CalculatorFactory()
         fitter = MultiFitter(model)  # default minimizer is LMFit, not BUMPS
 
-        core_called = {'count': 0}
-
-        def _should_not_be_called(**_kwargs):
-            core_called['count'] += 1
-            return {'draws': np.empty((0, 0)), 'param_names': [], 'state': None, 'logp': None}
-
-        fitter.easy_science_multi_fitter.mcmc_sample = _should_not_be_called
+        capture = {}
 
         data = sc.DataGroup({
             'coords': {'Qz_0': sc.array(dims=['Qz_0'], values=np.linspace(0.01, 0.3, 10))},
             'data': {'R_0': sc.array(dims=['Qz_0'], values=np.ones(10), variances=np.ones(10) * 0.01)},
         })
 
-        with pytest.raises(RuntimeError, match='Bayesian sampling requires a BUMPS minimizer'):
-            fitter.mcmc_sample(data)
-        assert core_called['count'] == 0
+        with _patch_sampler(capture) as sampler_cls:
+            with pytest.raises(RuntimeError, match='Bayesian sampling requires a BUMPS minimizer'):
+                fitter.mcmc_sample(data)
+        sampler_cls.assert_not_called()
 
 
 class TestMCMCSampleBasic:
     """Basic mcmc_sample() dispatch and return-value forwarding."""
 
-    def test_returns_core_result_dict(self):
-        """mcmc_sample() returns whatever the core MultiFitter.mcmc_sample() returns."""
+    def test_returns_result_dict_from_sampler(self):
+        """mcmc_sample() returns a dict built from the core Sampler's SamplingResults."""
         model = Model()
         model.interface = CalculatorFactory()
         fitter = MultiFitter(model)
 
-        # Mock the core MultiFitter.mcmc_sample to return a known dict
-        fake_result = {'draws': np.ones((10, 2)), 'param_names': ['a', 'b'], 'state': None, 'logp': None}
         fitter.easy_science_multi_fitter = MagicMock()
         fitter.easy_science_multi_fitter.minimizer.package = 'bumps'
-        fitter.easy_science_multi_fitter.mcmc_sample = MagicMock(return_value=fake_result)
+
+        draws = np.ones((10, 2))
+        sentinel_state = object()
+        logp = np.zeros(10)
+        results = _fake_sampling_results(draws=draws, param_names=['a', 'b'], state=sentinel_state, logp=logp)
+
+        capture = {}
 
         data = sc.DataGroup({
             'coords': {'Qz_0': sc.array(dims=['Qz_0'], values=np.linspace(0.01, 0.3, 10))},
             'data': {'R_0': sc.array(dims=['Qz_0'], values=np.ones(10), variances=np.ones(10) * 0.01)},
         })
 
-        result = fitter.mcmc_sample(data, samples=100, burn=20, thin=2, population=5)
-        assert result is fake_result
+        with _patch_sampler(capture, results=results):
+            result = fitter.mcmc_sample(data, samples=100, burn=20, thin=2, population=5)
+
+        # The fitter passed to Sampler is the core MultiFitter
+        assert capture['fitter'] is fitter.easy_science_multi_fitter
+        assert result['draws'] is draws
+        assert result['param_names'] == ['a', 'b']
+        assert result['state'] is sentinel_state
+        assert result['logp'] is logp
 
-    def test_forwards_hyperparams_to_core(self):
-        """Samples, burn, thin, population, chains are forwarded to core."""
+    def test_forwards_hyperparams_to_sampler(self):
+        """Samples, burn, thin, population are forwarded to Sampler.sample()."""
         model = Model()
         model.interface = CalculatorFactory()
         fitter = MultiFitter(model)
 
-        captured = {}
-
-        def _fake_mcmc_sample(*, x, y, weights, samples, burn, thin, population, **kwargs):
-            captured['samples'] = samples
-            captured['burn'] = burn
-            captured['thin'] = thin
-            captured['population'] = population
-            return {'draws': np.ones((10, 2)), 'param_names': ['a', 'b'], 'state': None, 'logp': None}
-
         fitter.easy_science_multi_fitter = MagicMock()
         fitter.easy_science_multi_fitter.minimizer.package = 'bumps'
-        fitter.easy_science_multi_fitter.mcmc_sample = MagicMock(side_effect=_fake_mcmc_sample)
+
+        capture = {}
 
         data = sc.DataGroup({
             'coords': {'Qz_0': sc.array(dims=['Qz_0'], values=np.linspace(0.01, 0.3, 10))},
             'data': {'R_0': sc.array(dims=['Qz_0'], values=np.ones(10), variances=np.ones(10) * 0.01)},
         })
 
-        fitter.mcmc_sample(data, samples=500, burn=100, thin=5, population=8)
-        assert captured['samples'] == 500
-        assert captured['burn'] == 100
-        assert captured['thin'] == 5
-        assert captured['population'] == 8
+        with _patch_sampler(capture):
+            fitter.mcmc_sample(data, samples=500, burn=100, thin=5, population=8)
+        assert capture['samples'] == 500
+        assert capture['burn'] == 100
+        assert capture['thin'] == 5
+        assert capture['population'] == 8
 
-    def test_forwards_population_to_core(self):
-        """'population' argument is forwarded to core."""
+    def test_forwards_population_to_sampler(self):
+        """'population' argument is forwarded to Sampler.sample()."""
         model = Model()
         model.interface = CalculatorFactory()
         fitter = MultiFitter(model)
 
-        captured = {}
-
-        def _fake_mcmc_sample(*, x, y, weights, population, **kwargs):
-            captured['population'] = population
-            return {'draws': np.ones((10, 2)), 'param_names': ['a', 'b'], 'state': None, 'logp': None}
-
         fitter.easy_science_multi_fitter = MagicMock()
         fitter.easy_science_multi_fitter.minimizer.package = 'bumps'
-        fitter.easy_science_multi_fitter.mcmc_sample = MagicMock(side_effect=_fake_mcmc_sample)
+
+        capture = {}
 
         data = sc.DataGroup({
             'coords': {'Qz_0': sc.array(dims=['Qz_0'], values=np.linspace(0.01, 0.3, 10))},
             'data': {'R_0': sc.array(dims=['Qz_0'], values=np.ones(10), variances=np.ones(10) * 0.01)},
         })
 
-        fitter.mcmc_sample(data, samples=100, burn=20, thin=2, population=6)
-        assert captured['population'] == 6
+        with _patch_sampler(capture):
+            fitter.mcmc_sample(data, samples=100, burn=20, thin=2, population=6)
+        assert capture['population'] == 6
 
 
 class TestMCMCSampleInitializer:
     """initializer parameter is forwarded via sampler_kwargs."""
 
     def test_initializer_passed_as_sampler_kwargs_init(self):
-        """initializer='lhs' should be passed as sampler_kwargs={'init': 'lhs'} to core."""
+        """initializer='lhs' should be passed as sampler_kwargs={'init': 'lhs'} to Sampler.sample()."""
         model = Model()
         model.interface = CalculatorFactory()
         fitter = MultiFitter(model)
 
-        captured = {}
-
-        def _fake_mcmc_sample(*, sampler_kwargs, **kwargs):
-            captured['sampler_kwargs'] = sampler_kwargs
-            return {'draws': np.ones((10, 2)), 'param_names': ['a', 'b'], 'state': None, 'logp': None}
-
         fitter.easy_science_multi_fitter = MagicMock()
         fitter.easy_science_multi_fitter.minimizer.package = 'bumps'
-        fitter.easy_science_multi_fitter.mcmc_sample = MagicMock(side_effect=_fake_mcmc_sample)
+
+        capture = {}
 
         data = sc.DataGroup({
             'coords': {'Qz_0': sc.array(dims=['Qz_0'], values=np.linspace(0.01, 0.3, 10))},
             'data': {'R_0': sc.array(dims=['Qz_0'], values=np.ones(10), variances=np.ones(10) * 0.01)},
         })
 
-        fitter.mcmc_sample(data, samples=100, burn=20, thin=2, initializer='lhs')
-        assert captured['sampler_kwargs'] == {'init': 'lhs'}
+        with _patch_sampler(capture):
+            fitter.mcmc_sample(data, samples=100, burn=20, thin=2, initializer='lhs')
+        assert capture['sampler_kwargs'] == {'init': 'lhs'}
 
     def test_initializer_none_omits_sampler_kwargs(self):
         """When initializer is None, sampler_kwargs should be None, not an empty dict."""
@@ -963,23 +993,19 @@ def test_initializer_none_omits_sampler_kwargs(self):
         model.interface = CalculatorFactory()
         fitter = MultiFitter(model)
 
-        captured = {}
-
-        def _fake_mcmc_sample(*, sampler_kwargs, **kwargs):
-            captured['sampler_kwargs'] = sampler_kwargs
-            return {'draws': np.ones((10, 2)), 'param_names': ['a', 'b'], 'state': None, 'logp': None}
-
         fitter.easy_science_multi_fitter = MagicMock()
         fitter.easy_science_multi_fitter.minimizer.package = 'bumps'
-        fitter.easy_science_multi_fitter.mcmc_sample = MagicMock(side_effect=_fake_mcmc_sample)
+
+        capture = {}
 
         data = sc.DataGroup({
             'coords': {'Qz_0': sc.array(dims=['Qz_0'], values=np.linspace(0.01, 0.3, 10))},
             'data': {'R_0': sc.array(dims=['Qz_0'], values=np.ones(10), variances=np.ones(10) * 0.01)},
         })
 
-        fitter.mcmc_sample(data, samples=100, burn=20, thin=2)
-        assert captured['sampler_kwargs'] is None
+        with _patch_sampler(capture):
+            fitter.mcmc_sample(data, samples=100, burn=20, thin=2)
+        assert capture['sampler_kwargs'] is None
 
 
 class TestMCMCSampleZeroVariance:
@@ -994,17 +1020,10 @@ def test_hybrid_transforms_zero_variance_points(self):
         # Use legacy_mask so zero-variance points are dropped
         fitter = MultiFitter(model, objective='legacy_mask')
 
-        captured = {}
-
-        def _fake_mcmc_sample(*, x, y, weights, **kwargs):
-            captured['x'] = x
-            captured['y'] = y
-            captured['weights'] = weights
-            return {'draws': np.ones((10, 2)), 'param_names': ['a', 'b'], 'state': None, 'logp': None}
+        capture = {}
 
         fitter.easy_science_multi_fitter = MagicMock()
         fitter.easy_science_multi_fitter.minimizer.package = 'bumps'
-        fitter.easy_science_multi_fitter.mcmc_sample = MagicMock(side_effect=_fake_mcmc_sample)
 
         qz = np.linspace(0.01, 0.3, 10)
         r = np.exp(-qz * 50)
@@ -1018,12 +1037,13 @@ def _fake_mcmc_sample(*, x, y, weights, **kwargs):
 
         with warnings.catch_warnings(record=True) as w:
             warnings.simplefilter('always')
-            fitter.mcmc_sample(data, samples=100, burn=20, thin=2)
+            with _patch_sampler(capture):
+                fitter.mcmc_sample(data, samples=100, burn=20, thin=2)
 
         # legacy_mask should drop the 2 zero-variance points
-        assert len(captured['x'][0]) == 8
-        assert len(captured['y'][0]) == 8
-        assert len(captured['weights'][0]) == 8
+        assert len(capture['x'][0]) == 8
+        assert len(capture['y'][0]) == 8
+        assert len(capture['weights'][0]) == 8
 
         mask_warnings = [str(ww.message) for ww in w if 'Masked' in str(ww.message)]
         assert len(mask_warnings) == 1
@@ -1037,16 +1057,10 @@ def test_per_call_objective_override(self):
         model.interface = CalculatorFactory()
         fitter = MultiFitter(model, objective='legacy_mask')  # default
 
-        captured = {}
-
-        def _fake_mcmc_sample(*, x, y, weights, **kwargs):
-            captured['x'] = x
-            captured['y'] = y
-            return {'draws': np.ones((10, 2)), 'param_names': ['a', 'b'], 'state': None, 'logp': None}
+        capture = {}
 
         fitter.easy_science_multi_fitter = MagicMock()
         fitter.easy_science_multi_fitter.minimizer.package = 'bumps'
-        fitter.easy_science_multi_fitter.mcmc_sample = MagicMock(side_effect=_fake_mcmc_sample)
 
         qz = np.linspace(0.01, 0.3, 10)
         r = np.exp(-qz * 50)
@@ -1061,6 +1075,7 @@ def _fake_mcmc_sample(*, x, y, weights, **kwargs):
         # Override to hybrid — should keep all 10 points
         with warnings.catch_warnings(record=True):
             warnings.simplefilter('always')
-            fitter.mcmc_sample(data, samples=100, burn=20, thin=2, objective='hybrid')
+            with _patch_sampler(capture):
+                fitter.mcmc_sample(data, samples=100, burn=20, thin=2, objective='hybrid')
 
-        assert len(captured['x'][0]) == 10  # all points kept (Mighell-substituted)
+        assert len(capture['x'][0]) == 10  # all points kept (Mighell-substituted)