Rebalance Rest Python groups within the existing runner budget

Makefile

@@ -32,18 +32,19 @@ REST_SPM_TESTS := $(TESTS)/core/test_spm_policy_family.py \
 # modules, then core/, then policy/, and remaining takes every other file.
 # code_health/test_rest_python_groups.py checks that the groups together
 # collect each file of the old single process exactly once.
-# Test time per group, from run 37198184683 (1,453 s in all) plus the three
-# core modules added since, timed in runs 37171744596, 37196338557 and
-# 37198179476 and scaled to the speed of 37198184683:
-#   cached-arrays (403 s): test_formulas_do_not_write_into_cached_arrays.
-#   heavy-a (350 s): md_poverty_line_credit_invariants (134 s),
-#     ald_determinism (94 s), dependent_net_investment_income_invariants
-#     (64 s), ssi_state_supplement_medicaid_dependency (59 s).
-#   heavy-b (345 s): marginal_tax_rate_coverage (124 s),
-#     multi_year_simulation (88 s), behavioral_response_measurements (84 s),
-#     dependent_losses_agi_invariants (49 s).
+# CI run 38141803955 measured RSS/wall: cached-arrays 3.3 GB/5:26,
+# heavy-a 8.1 GB/9:55 (including pending #10058), heavy-b 8.5 GB/7:31,
+# core 13.3 GB/11:51, policy 13.9 GB/8:46, remaining 14.4 GB/22:18.
+# JUnit measures per-file elapsed time, not memory. The extra boundaries
+# below isolate retained full models and target <= ~10 GB per process;
+# verify those estimates against each group's GNU time report in CI.
 #   core (142 s), policy (185 s), remaining (239 s).
-REST_PYTHON_GROUPS := cached-arrays heavy-a heavy-b core policy remaining
+# SSI resource deeming retains a reformed system at import and builds another
+# in its grid fixture. Core reached 13.3 GB in CI run 38141803955 after this
+# and three smaller modules arrived; isolate SSI rather than filling heavy-a
+# (8.1 GB including the pending capital-gains MTR change). These groups remain
+# sequential subprocesses on the existing Rest runner.
+REST_PYTHON_GROUPS := cached-arrays heavy-a heavy-b core-heavy core policy-parameters policy remaining-heavy remaining
 REST_CACHED_ARRAY_TESTS := $(TESTS)/test_formulas_do_not_write_into_cached_arrays.py
 REST_HEAVY_A_TESTS := $(TESTS)/test_md_poverty_line_credit_invariants.py \
 	$(TESTS)/core/test_ald_determinism.py \
@@ -53,7 +54,27 @@ REST_HEAVY_B_TESTS := $(TESTS)/core/test_marginal_tax_rate_coverage.py \
 	$(TESTS)/core/test_multi_year_simulation.py \
 	$(TESTS)/core/test_behavioral_response_measurements.py \
 	$(TESTS)/core/test_dependent_losses_agi_invariants.py
-REST_LISTED_TESTS := $(REST_CACHED_ARRAY_TESTS) $(REST_HEAVY_A_TESTS) $(REST_HEAVY_B_TESTS)
+# SSI runs before the federal-credit fixture and NY/AL pinned-system cache
+# are built. Pins have three bounded entries; keep them apart from the NIIT,
+# CTC and Medicaid systems below, which stay resident after their files end.
+REST_CORE_HEAVY_TESTS := $(TESTS)/core/test_ssi_resource_deeming_invariants.py \
+	$(TESTS)/test_federal_credit_limit_order.py \
+	$(TESTS)/test_pinned_branch_credit_dependents.py
+# These three parameter modules construct full reform trees; one also loads
+# a default Microsimulation only to read parameters. Release their heap before
+# the remaining policy tests, rather than stacking both in the 13.9 GB group.
+REST_POLICY_PARAMETER_TESTS := $(TESTS)/policy/baseline/parameters
+# The first four files retain three independent full systems (the two CTC
+# files share one). The last two use transient TX simulations and balance
+# elapsed time without warming the NY/AL pinned-system cache.
+REST_REMAINING_HEAVY_TESTS := $(TESTS)/test_niit_section_911_magi.py \
+	$(TESTS)/test_ctc_foreign_earned_income_exclusion.py \
+	$(TESTS)/test_ctc_foreign_earned_income_exclusion_property.py \
+	$(TESTS)/unit/test_medicaid_abd_ssi_rate_standard_properties.py \
+	$(TESTS)/test_capital_gain_distributions_loss_limit.py \
+	$(TESTS)/test_schedule_d_filer_net_capital_gain.py
+REST_LISTED_TESTS := $(REST_CACHED_ARRAY_TESTS) $(REST_HEAVY_A_TESTS) $(REST_HEAVY_B_TESTS) \
+	$(REST_CORE_HEAVY_TESTS) $(REST_POLICY_PARAMETER_TESTS) $(REST_REMAINING_HEAVY_TESTS)
 REST_IGNORES := $(addprefix --ignore=,$(TESTS)/policy/contrib $(TESTS)/microsimulation $(REST_SPM_TESTS))
 # $(call rest_pytest,<group>,<paths>,<more paths to ignore>) is the pytest
 # command for one group. CI sets REST_REPORT_DIR, and each group then writes
@@ -271,10 +292,16 @@ test-other-python-rest-heavy-a:
 	$(call rest_pytest,heavy-a,$(REST_HEAVY_A_TESTS))
 test-other-python-rest-heavy-b:
 	$(call rest_pytest,heavy-b,$(REST_HEAVY_B_TESTS))
+test-other-python-rest-core-heavy:
+	$(call rest_pytest,core-heavy,$(REST_CORE_HEAVY_TESTS))
 test-other-python-rest-core:
 	$(call rest_pytest,core,$(TESTS)/core,$(REST_LISTED_TESTS))
+test-other-python-rest-policy-parameters:
+	$(call rest_pytest,policy-parameters,$(REST_POLICY_PARAMETER_TESTS))
 test-other-python-rest-policy:
 	$(call rest_pytest,policy,$(TESTS)/policy,$(REST_LISTED_TESTS))
+test-other-python-rest-remaining-heavy:
+	$(call rest_pytest,remaining-heavy,$(REST_REMAINING_HEAVY_TESTS))
 test-other-python-rest-remaining:
 	$(call rest_pytest,remaining,$(TESTS)/,$(REST_LISTED_TESTS) $(TESTS)/core $(TESTS)/policy)
 test-microsimulation:

changelog.d/rebalance-rest-core-group.changed.md

@@ -0,0 +1 @@
+Isolate memory-heavy Rest Python tests in sequential groups on the existing CI runner.

policyengine_us/tests/code_health/test_rest_python_groups.py

@@ -312,17 +312,15 @@ def test_test_files_keep_distinct_import_names(reference):
 """
 
 
-@pytest.mark.parametrize(
-    "outcomes",
-    [
-        ["0", "1", "0", "killed", "0", "0"],
-        ["killed", "1", "1", "1", "1", "1"],
-        ["0"] * 6,
-    ],
-    ids=["one-fails-one-killed", "all-fail", "all-pass"],
-)
-def test_every_group_runs_and_any_failure_fails_the_target(tmp_path, groups, outcomes):
-    assert len(groups) == len(outcomes), "give each group one outcome"
+@pytest.mark.parametrize("scenario", ["one-fails-one-killed", "all-fail", "all-pass"])
+def test_every_group_runs_and_any_failure_fails_the_target(tmp_path, groups, scenario):
+    outcomes = ["0"] * len(groups)
+    if scenario == "one-fails-one-killed":
+        assert len(groups) >= 4, "need groups before and after the failures"
+        outcomes[1] = "1"
+        outcomes[3] = "killed"
+    elif scenario == "all-fail":
+        outcomes = ["killed"] + ["1"] * (len(groups) - 1)
     stub = tmp_path / "bin" / "pytest"
     stub.parent.mkdir()
     stub.write_text(STUB_PYTEST)