Summary
Add a crash-loop guard around LSP semantic enrichment so Gortex stops reconnecting to a provider that repeatedly exits during a single enrichment pass.
Use case
Issue #220 showed clangd crashing repeatedly during enrichment when repo clang-tidy checks were active. The immediate fix is to avoid that trigger for clangd defaults, but Gortex should also have a generic safety valve for any LSP server that crashes repeatedly.
Proposed behavior
If an LSP provider exits too many times during one enrichment pass, mark that provider's enrichment as failed or abandoned for the repo, stop scheduling more requests to it, log the provider name, workspace, reconnect count, and final error, and leave already-confirmed graph updates intact.
Notes
Per CONTRIBUTING.md, this is an enhancement request describing the feature and use case. A future implementation should include focused tests for repeated server exits and verify that transient one-off exits can still reconnect successfully.
Summary
Add a crash-loop guard around LSP semantic enrichment so Gortex stops reconnecting to a provider that repeatedly exits during a single enrichment pass.
Use case
Issue #220 showed clangd crashing repeatedly during enrichment when repo clang-tidy checks were active. The immediate fix is to avoid that trigger for clangd defaults, but Gortex should also have a generic safety valve for any LSP server that crashes repeatedly.
Proposed behavior
If an LSP provider exits too many times during one enrichment pass, mark that provider's enrichment as failed or abandoned for the repo, stop scheduling more requests to it, log the provider name, workspace, reconnect count, and final error, and leave already-confirmed graph updates intact.
Notes
Per CONTRIBUTING.md, this is an enhancement request describing the feature and use case. A future implementation should include focused tests for repeated server exits and verify that transient one-off exits can still reconnect successfully.