Roadmap¶

Done¶

TCP probe — connect check with a per-probe timeout.
HTTP probe — configurable path, method and timeout; any-2xx-by-default or an explicit accepted status set; redirects deliberately not followed; latency reported even on failure; a fresh keep-alive-disabled transport per probe.
Runner — one-shot initial delay, immediate first attempt, fixed period, success/failure thresholds rolling into Healthy / Unhealthy / Unknown, a buffered non-blocking Result stream, a goroutine-safe Status(), and an idempotent Stop().
IsTimeout — separates "never reached the target" from "the target refused".
Injectable clock for deterministic tests.
100% test coverage across the six 64-bit Go targets (amd64, arm64, riscv64, loong64, ppc64le, s390x), -race, gofmt + go vet clean.

TypeExec probe. Running a command inside the target (rather than a network check against it) needs an in-target agent channel this stateless network runner has no access to. New returns an explicit error for TypeExec today so a misconfiguration surfaces immediately, rather than silently doing nothing. It stays out of scope until such a channel exists.
Acting on the verdict. Restarting, rerouting, alerting — deliberately not here. The library reports Healthy / Unhealthy; the decision of what to do is the caller's. This split is what keeps the runner small, dependency-free and reusable.

Additional probe kinds that remain pure network checks (e.g. gRPC health, TLS-handshake-only) — anything expressible as a stateless attempt with no in-target agent.
Optional structured logging / metrics hooks, kept behind an interface so the core stays dependency-free.