PGPool default 250m CPU request was causing scheduling failures on the
single-node COULOMBCORE cluster (2 vCPU, 98% allocated). Reduced to
100m request / 200m limit — safe for a lightweight connection pooler.
See INC-001: the-custodian/ops/incidents/2026-03-25-gitea-pgpool-crashloop.md
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
