diff --git a/.gitea/workflows/ci-deploy.yml b/.gitea/workflows/ci-deploy.yml
index 03f252b..c3aa4bd 100644
--- a/.gitea/workflows/ci-deploy.yml
+++ b/.gitea/workflows/ci-deploy.yml
@@ -62,30 +62,10 @@ jobs:
       - name: Test backend (host smoke)
         run: dotnet test JobTrackerApi.Tests/JobTrackerApi.Tests.csproj --configuration Release --no-build --filter "FullyQualifiedName~CvBuilderTests.Every_catalog_theme_renders_valid_html"
 
-      # Runner-only failure, localised to the AiWorkspace tests by bisecting across steps (the job
-      # log is not readable via the Gitea API). Those 10 tests pass on Windows, in a clean Linux
-      # container, under a 1GB memory cap, with a custom-dir SDK and no DOTNET_ROOT, serially, and
-      # under a hostile locale/timezone -- so the trigger is specific to this runner host.
-      # See docs/infrastructure/runner-investigation.md. Collection parallelism stays off for determinism, the
-      # same reason the frontend runs --runInBand. Every test still runs; nothing is filtered.
-      # Post-reboot the failure is identical (smoke 1s pass, suite 3s fail), so it is deterministic,
-      # not an OOM/flake -- and host telemetry shows no disk/memory/PID pressure. The passing smoke is
-      # a PURE test (no EF, no Moq); every class known to fail uses EF InMemory + Moq, whose Castle
-      # DynamicProxy emits IL at runtime. A container seccomp/W^X policy that blocks runtime code
-      # generation would kill the host exactly this way. These ordered steps test that: pure classes
-      # first, then a Moq+EF class. Diagnostic scaffolding; no test is skipped or weakened.
-      - name: T pure JobPipeline
-        run: dotnet test JobTrackerApi.Tests/JobTrackerApi.Tests.csproj --configuration Release --no-build --filter "FullyQualifiedName~JobPipelineTests"
-
-      - name: T pure EmailStatusClassifier
-        run: dotnet test JobTrackerApi.Tests/JobTrackerApi.Tests.csproj --configuration Release --no-build --filter "FullyQualifiedName~EmailStatusClassifierTests"
-
-      - name: T pure JobCvMatchService
-        run: dotnet test JobTrackerApi.Tests/JobTrackerApi.Tests.csproj --configuration Release --no-build --filter "FullyQualifiedName~JobCvMatchServiceTests"
-
-      - name: T MoqEF CvBuilder
-        run: dotnet test JobTrackerApi.Tests/JobTrackerApi.Tests.csproj --configuration Release --no-build --filter "FullyQualifiedName~CvBuilderTests"
-
+      # This runner fails at a DIFFERENT stage on different runs with no readable diagnostics
+      # (test host at 3s; on another run `dotnet restore` at 0s). See
+      # docs/infrastructure/runner-investigation.md. Collection parallelism stays off for
+      # determinism, the same reason the frontend runs --runInBand. Every test runs; nothing skipped.
       - name: Test backend
         run: dotnet test JobTrackerApi.Tests/JobTrackerApi.Tests.csproj --configuration Release --no-build -- xUnit.parallelizeTestCollections=false xUnit.maxParallelThreads=1
 
diff --git a/docs/infrastructure/runner-investigation.md b/docs/infrastructure/runner-investigation.md
index 0e9cca4..a48f8a1 100644
--- a/docs/infrastructure/runner-investigation.md
+++ b/docs/infrastructure/runner-investigation.md
@@ -103,15 +103,41 @@ measured yet — see the asks below.
 The one host-level value worth noting is `ulimit -n 1024` (file descriptors), which is low by modern
 standards, though it is the interactive shell's value and not necessarily the runner service's.
 
+### Post-reboot runs — the failure moves between stages
+
+Two further runs after the machine was restarted:
+
+| Run | Result |
+|---|---|
+| `8f73548` (post-reboot) | **Identical** to pre-reboot: host smoke ✓ 1 s, full suite ✗ 3 s |
+| `c4c0cd4` | Failure **moved earlier**: `Restore backend tests` ✗ **0 s** — a step that took 3–4 s and succeeded in all previous runs. None of the queued diagnostic steps executed. |
+
+Two conclusions:
+
+1. **The reboot changed nothing**, so this is not stuck state, a leaked process, or a corrupted
+   workspace that a restart would clear.
+2. **The failing stage is not stable across runs.** `dotnet restore` failing in 0 s on one run and
+   succeeding in 3–4 s on the next, with the same commit and the same runner, is nondeterministic
+   infrastructure behaviour. It also argues *against* a deterministic explanation such as a seccomp /
+   W^X policy blocking runtime IL emission (which would fail identically every time), and against any
+   single-test explanation.
+
+Taken with the three failure modes the workflow already documents on this runner (SDK tar corruption,
+`npm ci` SIGSEGV, silent CRA build death), the pattern is a runner that intermittently kills or fails
+child processes at arbitrary stages, with no diagnostic surfaced.
+
 ### Confidence level
 
 - **Application code is not the cause — high confidence (~95 %).** Ten independent environments,
   including a byte-exact clean checkout, all pass. Every axis raised (Linux behaviour, case
   sensitivity, path separators, locale, time zone, environment variables, parallel execution, test
   ordering, shared state, memory) has been experimentally eliminated.
-- **Specific mechanism (OOM vs pids vs disk) — low/medium confidence (~40 %).** Not reproducible
-  remotely, and not resolvable without the job log or host access. I am deliberately not asserting
-  which one it is.
+- **Specific mechanism — low confidence (~25 %), and lower than before.** Host telemetry ruled out
+  disk, inode, memory and PID exhaustion; the reboot ruled out stuck state; the moving failure stage
+  ruled out a deterministic sandbox policy. What remains is an intermittent fault in the runner's
+  execution environment (most likely the job container / `act_runner` configuration rather than the
+  host), which cannot be identified without the job log or runner config. I am not asserting a
+  mechanism.
 
 ### Why application code is no longer suspected