[AHMETHAKANBEZIR1]

A GP returned by condition cannot predict or condition on another batch at new coordinates: its Conditioned mean stores include_mean as a dynamic PyTree leaf, then evaluates that traced Boolean in a Python if. The public predict and _condition JIT paths raise TracerBoolConversionError.

Mark this configuration flag static, consistent with the public conditioning methods' treatment of include_mean. Keep all mean/covariance calculations unchanged.

Validation on Windows / Python 3.12.14 / JAX 0.12.0 / Equinox 0.13.8, CPU:

• Four new regressions fail on unchanged main, covering both include-mean settings and both subsequent prediction and conditioning. All pass with this fix.
• New and existing GP tests: 14 passed.
• Full suite with default float32: 174 passed, 2 skipped; with JAX_ENABLE_X64=1: 174 passed, 2 skipped. Skips are optional george/celerite comparisons.
• Source doctests: 3 passed. Configured pre-commit hooks and git diff --check pass; a news fragment is included.

The tests compare two-stage posterior means, predictive variances, and conditional covariances with independently assembled dense Gaussian conditioning formulas.

Prepared autonomously with OpenAI Codex, including code and tests; no independent human review is claimed.