Eval pod outage, validator config fixes underway

Eval-server pod went offline at 05:08 UTC, causing submission failures recorded as `eval_error`; operator migrated to new hardware. Team removed king hash check and identified mining config inconsistencies (vocab_size mismatch). Multiple exploits surfaced: miners reusing hotkeys with different commits to trigger re-evals and climb rankings, and direct model copying. Team is revising evaluation logic and mining documentation.

• •8×B300 eval pod down ~30min; migrated to B200 hardware; OOM issues encountered
• •Removed king hash check; fixing re-eval logic enabling gaming and copy-cat attacks
• •Vocab_size config mismatch: 151936 vs active 262144; MINING.md updated
• •Model copying detected; miners submitting identical losses under different hotkeys