Tonight’s threshold is no longer “is there a patch?” but “is there evidence the system can still be trusted?” Cisco FMC leads the operational ranking because observed root-level tradecraft—web shells, reverse shells, proxies, command execution, or credential theft—turns patching into an isolation, preservation, rotation, and potentially rebuild decision. Artifactory follows closely: unknown administrators, administrative token:anonymous activity, exposed join keys, malicious plugins, shells, or repository modification imply both platform compromise and downstream artifact risk. GitLab exploitation remains operationally significant, although the evidence base presented here is narrower and confidence is therefore moderate. The Sandworm and Qilin labels should not change containment priorities; those attributions remain less certain than the exploitation itself.
For Mini Shai-Hulud, the strongest reported boundary is the TanStack publication chain: 84 malicious versions across 42 packages, linked to a pull_request_target workflow, cache poisoning, runner-memory access to an npm OIDC token, and publication through a legitimate trusted-publisher identity. The investigation must trace execution and publication authority—not merely package installation—while preserving caches, workflow evidence, OIDC claims, tarballs, and attestations. At the same time, Lena found no independent database support for broader victim scope, so plausible transitive exposure through shared runners, cloud credentials, or Kubernetes paths must remain an investigation hypothesis rather than a confirmed blast radius.
The financial cases also require disciplined reconciliation. Symbiosis’s enormous raw syBTC mint figure is not equivalent to realized loss; reported cash-out was about $336,000, while the claimed 15 BTC recovery remains unreconciled against unauthorized balances and counterparty liabilities. In Liquid, approximately 3,400 of roughly 4,000 withdrawn BTC was returned, leaving a disputed 598.5 BTC and unresolved claims despite completed bridge-node patching. Across all of these incidents, containment cannot be declared complete from patch status, recovery announcements, or attribution alone. The next step is to convert these evidence thresholds and uncertainties into a defensible architecture-level sequence for isolation, trust restoration, and return to service.