mirror: drain probe body before close so the idle pool works

[Soph]

https://entire.io/gh/entireio/cli/trails/474

Summary

probeClient (used by the mirror clone-readiness loop in repo_mirror_probe.go) is deliberately built with a small idle pool (MaxIdleConns: 2) and a 90s IdleConnTimeout, specifically to reuse the TLS session across the 2s probes. But Go's transport only returns a connection to the idle pool when the response body is read to EOF before Close().

mirrorAdvertisesHead closed the body without draining it, and the sr.Decode/adv.Decode error returns leave the body partially read. So the connection was closed instead of pooled — every 2s probe paid a fresh TLS handshake, defeating the pool's stated purpose. Over a 30-minute wait (~900 probes) that's 900 handshakes instead of one reused connection.

Change

Drain the body (bounded by maxProbeBytes to match the existing read cap) before closing, so the transport can recycle the connection.

Testing

• mise run fmt + mise run lint — clean
• go build ./cmd/entire/... — clean

🤖 Generated with Claude Code

---

Note

Low Risk
Localized HTTP client lifecycle fix in the mirror readiness probe; no auth, data, or API contract changes.

Overview
Fixes mirror clone-readiness probes so probeClient can actually reuse idle connections and TLS sessions between ~2s ticks.

In mirrorAdvertisesHead, the response body is now drained (capped at maxProbeBytes, same as the decode path) before Close(), because partial reads from decode/error paths prevented Go’s transport from returning connections to the pool—long waits could repeat a full TLS handshake on every probe instead of one reused connection.

^{Reviewed by Cursor Bugbot for commit 4aed46f. Configure here.}

[Copilot]

Pull request overview

Fixes connection reuse in the mirror clone-readiness probe loop. probeClient was configured with an idle pool and 90s IdleConnTimeout to reuse TLS sessions across ~2s probes, but mirrorAdvertisesHead closed the response body without draining it. Because Go's transport only returns connections to the idle pool when bodies are read to EOF before Close(), every probe was paying a fresh TLS handshake — defeating the pool entirely during long waits.

Changes:

• In mirrorAdvertisesHead, drain the response body (capped to maxProbeBytes to mirror the existing read cap) before Close() so the transport can recycle the connection.

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{❌ Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

Comment @cursor review or bugbot run to trigger another review on this PR

^{Reviewed by Cursor Bugbot for commit 4aed46f. Configure here.}