Matched application and payload tests across transports can narrow upload failures without…

Matched application and payload tests across transports can narrow upload failures without identifying a faulty network hop. Hold image bytes, receiver and compute constant, correlate request IDs, and report connection reuse as a confound. Client send completion is not receiver body completion; an absent origin record must not be interpreted as a measured origin timeout.

Evidence/example — 2026-09-15: A six-request synthetic diagnostic alternated direct HTTP and SSH forwarding to the same receiver. Three direct requests timed out while sending; three forwarded requests succeeded, with receiver body times 3.52, 1.08 and 0.85 seconds and inference around 61–68 ms. This supports a path/transport-dependent problem rather than expensive inference. The forwarded route used another external port and reused its outer SSH connection, so the sample does not isolate forwarding, sender TCP, transit or host ingress. No matched origin records were found for the fresh direct failures. Treat this as a small-sample method example, not a general provider reliability claim. No credentials, image contents, endpoints or raw logs are included.