test(compression): resolve context_length inside mocks across suites hit by lazy-init deferral

Baseline diff vs clean upstream/main (182 pre-existing environmental
failures on both sides) showed exactly 6 PR-caused failures, all the same
mechanism: tests construct ContextCompressor under a
get_model_context_length mock and read threshold_percent/threshold_tokens
after the with block, or build via object.__new__ and assign
context_length AFTER the derived budgets (the setter now resets the
lazily-cached budgets).

- test_compression_small_ctx_threshold_floor: resolve inside _make()
- test_cjk_token_estimation: resolve inside mock
- test_per_model_compression_threshold: resolve inside mock (2 tests)
- test_pre_compress_memory_context: assign context_length before
  threshold/tail/summary budgets; add summary_target_ratio

tests/agent/test_cjk_token_estimation.py

@@ -35,6 +35,8 @@ def test_cjk_tail_does_not_expand_to_english_char_budget():
             summary_target_ratio=0.2,
             quiet_mode=True,
         )
+        # Resolve while the mock is active (lazy init, #32221).
+        _ = compressor.context_length
 
     messages = [
         {"role": "user", "content": "head 1"},

tests/agent/test_compression_small_ctx_threshold_floor.py

@@ -20,9 +20,13 @@ from agent.context_compressor import ContextCompressor
 
 def _make(ctx: int, pct: float = 0.50) -> ContextCompressor:
     with patch.object(cc, "get_model_context_length", return_value=ctx):
-        return ContextCompressor(
+        comp = ContextCompressor(
             model="test/model", threshold_percent=pct, quiet_mode=True,
         )
+        # Resolve while the mock is active — lazy init (#32221) defers the
+        # window probe (and the floor application) past __init__.
+        _ = comp.context_length
+        return comp
 
 
 class TestSmallContextThresholdFloor:

tests/agent/test_pre_compress_memory_context.py

@@ -13,10 +13,14 @@ def _make_compressor():
     compressor = ContextCompressor.__new__(ContextCompressor)
     compressor.protect_first_n = 2
     compressor.protect_last_n = 5
-    compressor.tail_token_budget = 20_000
+    # Set context_length BEFORE the derived budgets: its setter resets the
+    # lazily-cached threshold/tail/summary budgets (#32221 lazy init), so
+    # assigning it later would clear the explicit values below.
     compressor.context_length = 200_000
     compressor.threshold_percent = 0.80
     compressor.threshold_tokens = 160_000
+    compressor.summary_target_ratio = 0.20
+    compressor.tail_token_budget = 20_000
     compressor.max_summary_tokens = 10_000
     compressor.quiet_mode = True
     compressor.compression_count = 0

tests/run_agent/test_per_model_compression_threshold.py

@@ -89,6 +89,8 @@ class TestContextCompressorModelThresholds:
             model_thresholds={"glm-5.2": 0.40},
             quiet_mode=True,
         )
+        # Resolve while mock is active (lazy init defers floor past __init__).
+        _ = cc.context_length
         # 256K < 512K → floor at 0.75; override 0.40 < 0.75, so floor wins
         assert cc.threshold_percent == 0.75
         assert cc.threshold_tokens == int(256_000 * 0.75)
@@ -113,6 +115,8 @@ class TestContextCompressorModelThresholds:
             threshold_percent=0.50,
             quiet_mode=True,
         )
+        # Resolve while mock is active (lazy init defers floor past __init__).
+        _ = cc.context_length
         # 256K < 512K → floored at 0.75
         assert cc.threshold_percent == 0.75
         assert cc.model_thresholds == {}