SWE Environment Issues, Sampling Moved to Targon

Affine experienced a day-long SWE environment outage with only 10 successful samples per hour. Root cause identified: dirty MEMORY data marking tasks finished prematurely (10–18 of 34 required events). Team migrated SWE and MEMORY sampling to a dedicated Affine Targon cluster (2×H200 per miner, max 8 quotas), keeping Chutes active during transition. SWE environment remains enabled despite miner calls to disable it pending fixes.

• •SWE sampling degraded to ~10 successful samples/hour; memory data corruption discovered
• •Migrated SWE and MEMORY sampling to Targon cluster; maintain Chutes during transition
• •SWE image pool near capacity (7,793/7,803); requires 300+ samples to prep remaining images
• •Team declined disabling SWE; new SWE-Frontier environment planned within weeks