Document deployment pull error handling strategy

[miru-agents]

Summary

Added clarifying documentation to the pull_deployments function explaining its error handling approach and the rationale behind it.

Key Changes

• Added a multi-line comment block documenting why pull_deployments intentionally short-circuits on the first per-deployment error rather than collecting errors and continuing
• Explained the critical safety concern: partial deployment pulls could cause removal of currently-deployed config instances before their replacements are fully in place
• Documented that aborting the entire pull preserves local state consistency and allows the next sync cycle to retry from scratch

Implementation Details

The comment clarifies an important distinction from pull_content_for_cfg_insts, which has different error handling semantics. This documentation ensures future maintainers understand the deliberate design choice to fail fast on deployment ingestion errors, protecting against inconsistent local state that could lead to unintended config instance removal during apply runs.

https://claude.ai/code/session_01LC5Gcb1hq4ArTRmHWKHvpg