Unverified50% confidenceFactExact time
llm-retry's with_failover feature silently retries 400 errors (such as prompt too long or JSON format errors) to the fallback model, masking the error and causing quality degradation that no one notices.
1
Sources
50%
Confidence
Medium-term (~90 days)
Relevance
10/19/2026
First Seen
Valid until: 1/17/2027
Sources
Related Entities
Related Claims
UnverifiedWhen an LLM node returns malformed JSON, the tool contract specifies not to retry indefinitely, but instead to log the error, skip that record, and mark the number of skipped records in the summary.76% similarUnverifiedWhen an agent repeatedly encounters 403 or 429 errors, standard tools throw them as recoverable exceptions, causing the LLM to retry repeatedly and trigger stricter blocking.75% similarUnverified切换到推理模型时因服务端严格校验tool_calls参数必须是合法JSON而返回400错误，一条格式错误的消息会毒化整个对话历史导致后续请求全部失败74% similarUnverifiedIn a failover architecture, a distinction should be made between errors worth retrying (429, 5xx, timeouts) and errors that should not be retried (400 parameter errors, 401 authentication failures).74% similarUnverifiedIn streaming LLM APIs, some relay stations silently insert data: [DONE] to end early, or stuff error messages such as {"error": "rate limit exceeded"} into JSON disguised as model answers.72% similar
Cite This Claim
Stable URI
https://kongchang.com/claim/1064752API
curl https://kongchang.com/api/v1/knowledge/claims/1064752MCP
get_claim(id=1064752)