Questions about Schema Validation ("JSON Schema Mode")

[prescod]

I'm reading this: https://docs.litellm.ai/docs/completion/json_mode

I have some questions:

1. Why does it imply that there is some special connection between Vertex and the enable_json_schema_validation option? Are we not supposed to use it with other providers?
2. Why does it default to off?
3. Why does it matter what provider i'm using? Couldn't LiteLLM detect when to turn it on or off?
4. Does LiteLLM do retries if the schema fails to validate client-side? If so, how many? Does it feed the errors back into the LLM to get the best fix?
5. Why is it a global variable instead of a completion option?