test(evals): keep the trace screencast out of failure evidence

A failed agent-chat.track-delegated-work attempt (15 min, live run
indicator on screen) produced two ~127 MB trace zips; the 256 MB private
allowance refused the whole bundle and no evidence survived. Thirteen
launch-1474 cells failed the same way. The bulk was the trace's own
JPEG screencast, which does not compress. Keep DOM snapshots, network
and sources in the trace and rely on the retained failure video for the
visual timeline.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

tests/runner-e2e/playwright.config.ts

@@ -62,7 +62,12 @@ export default defineConfig({
     actionTimeout: 30_000,
     navigationTimeout: 30_000,
     screenshot: process.env.PAPERCLIP_RUNNER_E2E_PUBLIC_MCP === "1" ? "off" : "only-on-failure",
-    trace: process.env.PAPERCLIP_RUNNER_E2E_PUBLIC_MCP === "1" ? "off" : "retain-on-failure",
+    // Keep DOM snapshots and network, but not the trace's own screencast: a
+    // long chat attempt with a live run indicator streamed thousands of JPEG
+    // frames (incompressible), pushing failure evidence past the 256 MB private
+    // allowance so no evidence survived. The retained failure video still
+    // records the visual timeline.
+    trace: process.env.PAPERCLIP_RUNNER_E2E_PUBLIC_MCP === "1" ? "off" : { mode: "retain-on-failure", screenshots: false },
     // A developer-supplied system Chromium keeps the local smoke loop
     // installation-free; CI's managed browser retains failure video as usual.
     video: chromiumExecutable || process.env.PAPERCLIP_RUNNER_E2E_PUBLIC_MCP === "1" ? "off" : "retain-on-failure",