Something Broke After Maintenance: Compare Before and After
Separate expected convergence from a real post-maintenance fault by using the change boundary and saved pre-checks.
Start with everything the change actually touched
Identify interfaces, routes, optics, VLANs, policies, cabling, software processes, and dependent services changed during the window. Do not assume the documented scope was the real scope until post-checks prove it.
Compare the first failing timestamp with implementation steps. The timing can point directly to the action that altered service.
Compare evidence, not memory
Use saved pre-checks, configuration snapshots, route tables, optical readings, and interface state. Under pressure, engineers do not reliably remember dozens of small details. A focused config diff can expose an unintended shutdown, policy change, neighbor, or interface parameter.
Separate normal routing convergence from persistent failure by using expected timers and service criteria instead of waiting indefinitely.
Rollback when rollback is the lower-risk diagnostic
If the change is the leading cause and the backout is known and safe, rollback can restore service while providing strong evidence. Do not continue troubleshooting forward simply because significant effort has already been invested.
After rollback, repeat the customer-facing checks and verify all redundant elements returned. A service can appear restored while the network remains single-threaded.