fix(memory): preserve Phase 3 binding teardown and selection

[justinchuby]

Summary

• drop shared-prefix resources before their binding/context pin
• restore a healthy prior mechanism when a raced selection candidate becomes inactive
• document Phase-3 drop-driven teardown and the mapped-attribution adoption boundary

Part of #1186; fixes rejected Phase-3 revision #1279

Validation

• new drop-order and failed-selection regressions pass in isolation; each failed under a controlled test-only inversion/revert (exit 101)
• cargo test -p onnx-runtime-memory-api: 25 passed
• cargo test -p onnx-runtime-memory-governor: 63 passed
• cargo test -p onnx-runtime-ep-cpu --test shared_allocator: 8 passed
• cargo test -p onnx-runtime-ep-cuda --lib: 382 passed, 22 ignored
• cargo test -p onnx-runtime-cuda-memory --lib: 8 passed
• cargo test -p onnx-genai-ort governed_allocator: 20 passed
• real registry stress: 20/20 runs, 4,800 iterations
• scoped Clippy and strict memory-API rustdoc pass
• workspace format, all-target check, and governor rustdoc reproduce exact-base inherited failures only

[codecov]

Codecov Report

❌ Patch coverage is 92.18750% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 80.36%. Comparing base (31a8f28) to head (5f1b249).

Files with missing lines	Patch %	Lines
crates/onnx-runtime-memory-api/src/binding.rs	92.18%	2 Missing and 3 partials ⚠️

Additional details and impacted files

@@                            Coverage Diff                             @@
##           justinchuby-1186-memory-binding-phase3    #1283      +/-   ##
==========================================================================
+ Coverage                                   79.87%   80.36%   +0.49%     
==========================================================================
  Files                                         363      363              
  Lines                                      159604   159665      +61     
  Branches                                   159604   159665      +61     
==========================================================================
+ Hits                                       127483   128315     +832     
+ Misses                                      27426    26649     -777     
- Partials                                     4695     4701       +6     

Flag	Coverage Δ
mlas	85.32% <ø> (+0.26%)	⬆️
offline	80.26% <92.18%> (+0.49%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Files with missing lines	Coverage Δ
crates/onnx-runtime-memory-api/src/binding.rs	64.04% <92.18%> (+2.44%)	⬆️

... and 9 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[github-actions]

🔴 Benchmark Regression Detected

Comparison of criterion micro-benchmarks: PR head vs merge-base, measured on the same runner in the same job (base first → PR second).

ℹ️ Absolute times are informational only — they vary with runner load. The % change column is the reliable signal because both sides ran under identical conditions.

Status	Scenario	Base	PR	Change
🔴	matmul/medium_generic_f32_threads=8/32x512x512	1.14 ms	1.79 ms	+57.2%
🔴	block_quantized_matmul_cached_dense/mxfp4_uncached_dequant_each_call/1x1024x1024	553.88 µs	860.96 µs	+55.4%
🔴	matmul/medium_generic_bf16_threads=8/32x512x512	379.47 µs	553.94 µs	+46.0%
🔴	matmul/medium_generic_f16_threads=1/32x512x512	32.64 µs	47.00 µs	+44.0%
⚠️	gather/large_f32_threads=1-internal/131072	22.63 µs	28.61 µs	+26.5%
⚠️	qwen3_sampling_processors/top_k_top_p_full_sort_baseline	5.28 ms	6.41 ms	+21.3%
⚠️	add/large_bf16_threads=1-internal/4194304	1.59 ms	1.84 ms	+16.0%
✅	sampling_latency/greedy_per_token	2.97 µs	3.22 µs	+8.4%
✅	matmul/medium_generic_bf16_threads=1/32x512x512	525.85 µs	569.23 µs	+8.2%
✅	add/small_f16_threads=1-internal/1024	418.9 ns	443.7 ns	+5.9%
✅	kv_cache/alloc_dealloc_pages	36.21 µs	37.87 µs	+4.6%
✅	qwen3_sampling_processors/top_k_full_sort_baseline	1.96 ms	2.04 ms	+4.4%
✅	matmul/medium_generic_f32_threads=1/32x512x512	2.52 ms	2.60 ms	+2.8%
✅	sampling_latency/top_k_per_token	48.60 µs	49.64 µs	+2.2%
✅	tokenization/decode_tokens_per_second	5.89 ms	6.00 ms	+2.0%
✅	matmul/large_generic_bf16_threads=1/32x1024x1024	2.09 ms	2.13 ms	+1.9%
✅	matmul/large_generic_f16_threads=8/32x1024x1024	94.47 µs	95.94 µs	+1.6%
✅	add/medium_f16_threads=1-internal/262144	97.09 µs	98.45 µs	+1.4%
✅	sampling_latency/top_p_per_token	352.65 µs	357.55 µs	+1.4%
✅	gather/medium_f32_threads=1-internal/32768	3.46 µs	3.50 µs	+1.2%
✅	grammar_masking/llguidance_compute_mask/32	69.99 µs	70.44 µs	+0.6%
✅	logit_processing/seven_processor_chain_per_step	297.67 µs	299.56 µs	+0.6%
✅	qwen3_sampling_processors/top_k_partial_selection	130.75 µs	131.20 µs	+0.3%
✅	gather/small_f32_threads=1-internal/4096	619.5 ns	621.3 ns	+0.3%
✅	matmul/medium_generic_f16_threads=8/32x512x512	38.26 µs	38.30 µs	+0.1%
✅	add/large_f16_threads=1-internal/4194304	1.56 ms	1.56 ms	-0.2%
✅	qwen3_sampling_processors/top_k_top_p_fast	611.31 µs	609.88 µs	-0.2%
✅	reduce_mean/medium_f32_threads=1-internal/65536	228.48 µs	227.54 µs	-0.4%
✅	gather/small_f16_threads=1-internal/4096	444.5 ns	441.9 ns	-0.6%
✅	add/small_bf16_threads=1-internal/1024	415.7 ns	412.0 ns	-0.9%
✅	matmul/small_generic_f32_threads=1/1x256x256	34.16 µs	33.85 µs	-0.9%
✅	add/medium_bf16_threads=1-internal/262144	97.00 µs	95.48 µs	-1.6%
✅	reduce_mean/small_f32_threads=1-internal/4096	14.32 µs	13.99 µs	-2.3%
✅	add/medium_f32_threads=1-internal/262144	22.85 µs	22.28 µs	-2.5%
✅	matmul/small_generic_f32_threads=8/1x256x256	34.65 µs	33.76 µs	-2.6%
✅	sampling_latency/min_p_per_token	202.57 µs	197.14 µs	-2.7%
✅	reduce_mean/large_f32_threads=1-internal/262144	942.88 µs	909.82 µs	-3.5%
✅	tokenization/encode_tokens_per_second	364.51 µs	349.57 µs	-4.1%
✅	gather/medium_f16_threads=1-internal/32768	2.31 µs	2.21 µs	-4.2%
✅	gather/small_bf16_threads=1-internal/4096	460.0 ns	439.8 ns	-4.4%
✅	add/large_f32_threads=1-internal/4194304	624.78 µs	590.84 µs	-5.4%
✅	matmul/large_generic_f32_threads=1/32x1024x1024	10.44 ms	9.82 ms	-6.0%
✅	qwen3_sampling_processors/top_p_full_sort_after_top_k_baseline	3.54 ms	3.29 ms	-7.2%
✅	matmul/large_generic_f16_threads=1/32x1024x1024	86.21 µs	78.87 µs	-8.5%
✅	matmul/large_generic_bf16_threads=8/32x1024x1024	1.74 ms	1.56 ms	-10.6%
✅	qwen3_sampling_processors/top_p_fast_after_top_k	546.51 µs	483.75 µs	-11.5%
✅	gather/medium_bf16_threads=1-internal/32768	2.56 µs	2.21 µs	-13.8%
✅	add/small_f32_threads=1-internal/1024	209.1 ns	179.7 ns	-14.1%
🟢	matmul/large_generic_f32_threads=8/32x1024x1024	4.94 ms	3.82 ms	-22.6%
🟢	matmul/small_generic_f16_threads=1/1x256x256	37.24 µs	28.38 µs	-23.8%
🟢	block_quantized_moe_cached_dense/mxfp4_cached_dense_expert_repeated_call/rows=1,H=256,I=256,E=4,top_k=1	89.50 µs	61.66 µs	-31.1%
🟢	matmul/small_generic_bf16_threads=1/1x256x256	48.23 µs	33.23 µs	-31.1%
🟢	gather/large_f16_threads=1-internal/131072	13.88 µs	9.38 µs	-32.4%
🟢	gather/large_bf16_threads=1-internal/131072	15.09 µs	10.00 µs	-33.7%
🟢	block_quantized_matmul_cached_dense/mxfp4_preexpanded_dense_oncelock_like_proxy/1x1024x1024	108.88 µs	71.21 µs	-34.6%
🟢	matmul/small_generic_bf16_threads=8/1x256x256	46.99 µs	30.03 µs	-36.1%
🟢	matmul/small_generic_f16_threads=8/1x256x256	45.55 µs	28.40 µs	-37.6%
🟢	block_quantized_moe_cached_dense/mxfp4_uncached_expert_dequant_each_call/rows=1,H=256,I=256,E=4,top_k=1	700.58 µs	365.62 µs	-47.8%
🟢	block_quantized_matmul_cached_dense/mxfp4_cached_dense_repeated_call/1x1024x1024	128.83 µs	46.28 µs	-64.1%

Visual flags: ⚠️ ≥ 15% slower, 🔴 ≥ 30% slower — calibrated against measured runner noise (~27% worst-case on multi-threaded matmul)

Host info

CPU: Apple M1 (Virtual)
Cores: 3
OS: Darwin 25.5.0 arm64
Rust: rustc 1.97.1 (8bab26f4f 2026-07-14)
Load avg: { 2.65 3.02 4.52 }

What this cannot catch

• Regressions in code paths not covered by these benchmarks (e.g., end-to-end decode with a real model)
• Sub-threshold regressions that compound over multiple PRs
• Performance changes that only manifest under GPU execution
• Latency changes in the ORT integration path (these benchmarks exercise the native Rust kernels)

[justinchuby]

Closing: superseded by #1579, which collapsed phases 1–7 into a single merge — landed as a36964280 (squash, 126 files, +43956 −3286).

This PR was approved by a session that was not its author, and its work is in main.

Containment verified, not assumed:

$ git merge-base --is-ancestor 5f1b249e2 ebceab071   # ebceab071 = #1579 head
→ ancestor

Every one of the 15 stack heads (#1252 → #1533) is a literal ancestor of the merged head, and I separately confirmed the merged head's content reached main: of the 126 files #1579 touched, exactly one differs from main — crates/onnx-runtime-session/src/executor/tests.rs, where main carries 120 extra lines from #1703 (Expand-broadcast mask tests). That is the conflict resolution correctly preserving the other side, not content loss.

Why this didn't close itself: #1579's body says Closes the stack: #1252 #1263 …, which GitHub does not parse — the closing keyword must be immediately followed by the reference, and in any case closing keywords only ever close issues, never pull requests. So all 15 stayed open by mechanism, not by intent.

#1186 stays open: #1579 explicitly gates it on memory-plugin-provider-wiring (the back half of Phase 6 criterion 8, provider/context pinning), which has not landed.