fix(approuter): bump memory 512M→2048M to stop prod OOM crash-loop

[jung-thomas]

Problem

The prod approuter tutorials-prod-approuter (serves developers.sap.com) crash-looped in production on 2026-08-10 with exit status 137 — container cgroup OOM kill at its 512M memory limit.

Root cause

• Fresh deploy at 10:40 EDT → developers.sap.com vanity route mapped at 13:10 EDT → OOM crash-loop 13:37–13:55 EDT once real prod traffic arrived.
• Exit 137 is a container-level cgroup OOM (total RSS > limit), distinct from the srv module's earlier exit-134 V8 heap SIGABRT — but the same class of under-provisioning. The srv module was already bumped 1G→2G for this reason (mta.yaml:96); the approuter was never given headroom.
• 512M was too little for concurrent request buffering + the large static-asset surface (admin / analytics / explore / scanner / data-inspector bundles copied into the approuter at build).

Fix

• tutorials-approuter memory: 512M → 2048M in .deploy/mta.yaml.
• MTA version 1.13.1 → 1.13.2 (patch — config fix).
• Already live in prod via emergency cf scale ... -m 2048M; this bakes it into the descriptor so the next deploy doesn't silently revert to 512M.

Disk

Checked per request: disk-quota is already 2048M and steady-state usage is ~33% (682M). Approuter logs stream off-container to stdout/stderr, so log growth is not a disk/OOM vector — no disk change needed.

Verification

• Emergency scale to 2G stopped the crash loop; app running since 17:56Z at ~261M / 2G (flat, not climbing → load pressure, not a leak).
• yq confirms descriptor parses and reports memory: 2048M, disk-quota: 2048M, version: 1.13.2.