TEEServiceRegistry has only 1 valid LLM executor, endpoint reachable but inference fails

Running integration work against the LLM precompile (0x0802) and found

TEEServiceRegistry currently has only 1 valid registered executor

(0xB42e435c4252A5a2E7440e37B609F00c61a0c91B). Its HTTP endpoint responds,

but any inference call fails with:

"Failed to get vLLM client: failed to get cert hash for endpoint from registry"

Flagged this to Jez in Discord (#build) and confirmed the team is aware.

Also saw another builder (Zegna) hit the identical error independently.

This blocks any work needing genuine multi-executor consensus (anything

requiring ≥3 distinct working executors). Happy to share a repro transaction

if useful.