diff --git a/bench/package.json b/bench/package.json
index b0e72a0..527b1b8 100644
--- a/bench/package.json
+++ b/bench/package.json
@@ -18,7 +18,7 @@
   "dependencies": {
     "@tangle-network/agent-eval": "^0.83.0",
     "@tangle-network/agent-runtime": "file:..",
-    "@tangle-network/sandbox": "^0.4.0"
+    "@tangle-network/sandbox": "^0.4.3"
   },
   "devDependencies": {
     "tsx": "^4.19.0",
diff --git a/bench/pnpm-lock.yaml b/bench/pnpm-lock.yaml
index 82bb57d..7d03d2d 100644
--- a/bench/pnpm-lock.yaml
+++ b/bench/pnpm-lock.yaml
@@ -10,13 +10,13 @@ importers:
     dependencies:
       '@tangle-network/agent-eval':
         specifier: ^0.83.0
-        version: 0.83.0(@tangle-network/sandbox@0.4.2(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))(typescript@5.9.3)
+        version: 0.83.0(@tangle-network/sandbox@0.4.3(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))(typescript@5.9.3)
       '@tangle-network/agent-runtime':
         specifier: file:..
-        version: file:..(@tangle-network/agent-eval@0.83.0(@tangle-network/sandbox@0.4.2(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))(typescript@5.9.3))(@tangle-network/sandbox@0.4.2(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))
+        version: file:..(@tangle-network/agent-eval@0.83.0(@tangle-network/sandbox@0.4.3(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))(typescript@5.9.3))(@tangle-network/sandbox@0.4.3(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))
       '@tangle-network/sandbox':
-        specifier: ^0.4.0
-        version: 0.4.2(viem@2.52.0(typescript@5.9.3)(zod@4.4.3))
+        specifier: ^0.4.3
+        version: 0.4.3(viem@2.52.0(typescript@5.9.3)(zod@4.4.3))
     devDependencies:
       tsx:
         specifier: ^4.19.0
@@ -268,7 +268,7 @@ packages:
     engines: {node: '>=20'}
     hasBin: true
     peerDependencies:
-      '@tangle-network/agent-eval': '>=0.76.0 <1.0.0'
+      '@tangle-network/agent-eval': '>=0.83.0 <1.0.0'
       '@tangle-network/agent-knowledge': '>=1.3.0 <2.0.0'
       '@tangle-network/sandbox': '>=0.1.2 <0.5.0'
       playwright: ^1.40.0
@@ -300,8 +300,8 @@ packages:
       viem:
         optional: true
 
-  '@tangle-network/sandbox@0.4.2':
-    resolution: {integrity: sha512-+S9Wnrdvc1pD1cVuu6GBVT9UE+p7FVmjCj9JWU06zvZu6bdcpf4ehMUaxoTgY9/3NlykCIlaWO4mrH89/k3/5Q==}
+  '@tangle-network/sandbox@0.4.3':
+    resolution: {integrity: sha512-6QE3Nuhkd8f+OlpRJbumHTAG4wKR+ESXT47UE0fjTf7ndRWLnhE4RZ7YRtHVo/Q9ZZr0FGH1mwM+6tW0NAT1bA==}
     peerDependencies:
       '@mastra/core': ^1.36.0
       '@modelcontextprotocol/sdk': ^1.29.0
@@ -558,7 +558,7 @@ snapshots:
       '@noble/hashes': 2.2.0
       '@scure/base': 2.2.0
 
-  '@tangle-network/agent-eval@0.83.0(@tangle-network/sandbox@0.4.2(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))(typescript@5.9.3)':
+  '@tangle-network/agent-eval@0.83.0(@tangle-network/sandbox@0.4.3(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))(typescript@5.9.3)':
     dependencies:
       '@asteasolutions/zod-to-openapi': 8.5.0(zod@4.4.3)
       '@ax-llm/ax': 19.0.45(zod@4.4.3)
@@ -567,7 +567,7 @@ snapshots:
       hono: 4.12.23
       zod: 4.4.3
     optionalDependencies:
-      '@tangle-network/sandbox': 0.4.2(viem@2.52.0(typescript@5.9.3)(zod@4.4.3))
+      '@tangle-network/sandbox': 0.4.3(viem@2.52.0(typescript@5.9.3)(zod@4.4.3))
     transitivePeerDependencies:
       - '@mastra/core'
       - '@modelcontextprotocol/sdk'
@@ -579,11 +579,11 @@ snapshots:
 
   '@tangle-network/agent-integrations@0.29.0': {}
 
-  '@tangle-network/agent-runtime@file:..(@tangle-network/agent-eval@0.83.0(@tangle-network/sandbox@0.4.2(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))(typescript@5.9.3))(@tangle-network/sandbox@0.4.2(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))':
+  '@tangle-network/agent-runtime@file:..(@tangle-network/agent-eval@0.83.0(@tangle-network/sandbox@0.4.3(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))(typescript@5.9.3))(@tangle-network/sandbox@0.4.3(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))':
     dependencies:
-      '@tangle-network/agent-eval': 0.83.0(@tangle-network/sandbox@0.4.2(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))(typescript@5.9.3)
+      '@tangle-network/agent-eval': 0.83.0(@tangle-network/sandbox@0.4.3(viem@2.52.0(typescript@5.9.3)(zod@4.4.3)))(typescript@5.9.3)
     optionalDependencies:
-      '@tangle-network/sandbox': 0.4.2(viem@2.52.0(typescript@5.9.3)(zod@4.4.3))
+      '@tangle-network/sandbox': 0.4.3(viem@2.52.0(typescript@5.9.3)(zod@4.4.3))
 
   '@tangle-network/sandbox@0.3.0(viem@2.52.0(typescript@5.9.3)(zod@4.4.3))':
     dependencies:
@@ -591,7 +591,7 @@ snapshots:
     optionalDependencies:
       viem: 2.52.0(typescript@5.9.3)(zod@4.4.3)
 
-  '@tangle-network/sandbox@0.4.2(viem@2.52.0(typescript@5.9.3)(zod@4.4.3))':
+  '@tangle-network/sandbox@0.4.3(viem@2.52.0(typescript@5.9.3)(zod@4.4.3))':
     optionalDependencies:
       viem: 2.52.0(typescript@5.9.3)(zod@4.4.3)
 
diff --git a/bench/src/experiment.ts b/bench/src/experiment.ts
index bc0db33..b645955 100644
--- a/bench/src/experiment.ts
+++ b/bench/src/experiment.ts
@@ -311,6 +311,10 @@ export async function runExperiment(cfg: ExperimentConfig): Promise<ExperimentRe
     })
     const iter0 = result.iterations[0]
     const infraError = iter0?.error !== undefined && iter0.output === undefined
+    if (infraError)
+      console.error(
+        `  [infra-cause] ${label} ${task.id}: ${(iter0?.error instanceof Error ? (iter0.error.stack ?? iter0.error.message) : String(iter0?.error)).slice(0, 700)}`,
+      )
     const resolved = result.winner?.verdict?.valid === true
     if (cfg.corpusPath) {
       // Fail-loud on a dropped row: a silent drop would leave the corpus with
diff --git a/src/runtime/sandbox-lineage.ts b/src/runtime/sandbox-lineage.ts
index aa6a9bd..634ca20 100644
--- a/src/runtime/sandbox-lineage.ts
+++ b/src/runtime/sandbox-lineage.ts
@@ -71,12 +71,17 @@ async function* pollPromptEvents(
   signal: AbortSignal,
 ): AsyncIterable<SandboxEvent> {
   if (signal.aborted) throwAbort()
-  await box.dispatchPrompt(prompt, { sessionId, signal })
-  const result = await box.session(sessionId).result()
+  // dispatchPrompt returns the session id the platform actually assigned, which
+  // may be one it MINTED rather than the supplied `sessionId`. Polling the
+  // supplied id when the platform minted a different one 404s the session-events
+  // endpoint ("Resource not found"). Always follow the assigned id.
+  const dispatched = await box.dispatchPrompt(prompt, { sessionId, signal })
+  const activeSessionId = dispatched.sessionId
+  const result = await box.session(activeSessionId).result()
   if (signal.aborted) throwAbort()
   yield {
     type: 'result',
-    id: sessionId,
+    id: activeSessionId,
     data: {
       finalText: result.response ?? '',
       success: result.success,