CVE-2026-14948 is a compound vulnerability where plaintext session tokens in accessible error log archives combine with weak access controls to allow low-privilege users to hijack administrator sessions. The CVSS score focuses on the exploit chain's severity, but the deeper failure is architectural: session identifiers were treated as debugging metadata rather than credential-equivalent values, and the log repository was served without verifying the caller's privilege level. What makes this different from a typical secret leak is the blast geometry — a single harvested admin session token doesn't just compromise one account, it collapses the privilege boundary entirely, often allowing credential resets, bulk data access, and access control modifications. The EPSS of 0.00392 is misleading as a rarity indicator; session tokens in logs is a normalized practice across far more systems than get reported, and this CVE likely represents one of the luckier breakages where the exposure happened to be exploitable rather than silently contained. The logging pipeline that introduced plaintext session tokens almost certainly predates the current threat model — it was probably reasonable debugging instrumentation when introduced, but was never re-evaluated as the system evolved. This is architectural drift: the threat model tightened around APIs and file endpoints, but the forgotten logging path kept serving artifacts under old assumptions. Organizations should audit their logging pipelines for any credential-adjacent values (session tokens, cookie headers, auth tokens) and assume that any privileged session active during the log retention window is potentially compromised if those logs are accessible to low-privilege users. The fix isn't just restricting archive access — it's reclassifying session tokens as secrets throughout the system, with logging pipeline hygiene that treats anything crossing privilege boundaries as a potential secret sink.