CVE-2026-62718 is an integer underflow in Microsoft's DHCP server that warrants more attention than its medium severity and low EPSS score suggest. The vulnerability exists in option parsing logic—specifically in how the server handles length values when processing DHCP options, typically in relay agent information or option 82 handling. The adjacent network attack vector is being sold as a meaningful constraint, but in modern enterprise environments, that boundary has collapsed. Wireless access points, VPN concentrators, and BYOD onboarding networks place attackers in an adjacent relationship to DHCP services by default. Every contractor on guest WiFi, every remote worker, every compromised laptop running reconnaissance—these are the realistic threat actors, not someone physically tapped onto a VLAN. The EPSS probability is measuring historical adjacent-network exploit rates, not current enterprise topology reality. But the deeper problem isn't the attack surface—it's the source code. Microsoft DHCP servers have accumulated option handlers across three decades of Windows Server releases. When option parsing logic gets refactored, the old parsers aren't deleted; they're marked internal, stripped from documentation, and left in the binary as fallback paths for backward compatibility. That's where integer underflows breed—in deprecated code paths that nobody on the current team knows why they're there, tested by nobody, and reviewed by no one because DHCP code review gets assigned to whoever has bandwidth, not whoever has threat model expertise. The patch likely addresses this specific underflow, but the pattern historyrhyme identifies suggests more are lurking in adjacent option parsing paths. For defenders: treat this as a forcing function to inventory which DHCP option handling paths are actually live versus deprecated. If your DHCP server has been in production for more than three major OS versions, you're running buried legacy code that likely has similar issues. Prioritize patching, but also plan for a code audit—the next integer underflow in this same server is probably waiting in an adjacent deprecated handler, and the EPSS score measures whether it will be exploited after disclosure, not whether it already has been.