[abelonogov-ld]

Requirements

• I have added test coverage for new or changed functionality
• I have followed the repository's pull request submission guidelines
• I have validated my changes against all supported platform versions

Related issues

Customer report: at cold boot, an app with seven environments opened 19 LaunchDarkly sockets. Seven were the expected flag streams. The other 12 carried 14 small POSTs to the events host that returned no body, at 349 ms p50 / 1,230 ms p90, which is mostly connection setup.

Describe the solution you've provided

Each environment's DarklyService created its own URLSession. Every URLSession has its own connection pool, so environments could never reuse each other's connections. Seven environments posting diagnostics and events to mobile.launchdarkly.com therefore needed at least seven TCP+TLS handshakes, even though that host negotiates HTTP/2 and one connection could carry all of them.

The session's configuration doesn't depend on the environment: minimum TLS 1.2 and no URL cache. Headers, the timeout and headerDelegate are set on each request. So DarklyService now uses a single shared session, DarklyService.session, and requests to the same host reuse connections across environments. The same applies to flag requests in polling mode.

Flag streams are unaffected: LDSwiftEventSource creates its own session for each stream.

Test: DarklyServiceSpec checks that services for two environments use the same session.

Describe alternatives you've considered

• Have ClientServiceFactory own the session and pass it to each service. That shares a session only within one LDClient.start, needs a change to the factory protocol and mock, and gains nothing, since the configuration never varies.
• Hold uploads back until all environments are ready, or batch them across environments. That is a bigger behavior change and doesn't address the per-session connection pools.

Additional context

Before this change the sessions were never invalidated either. A shared, process-lifetime session is the usual URLSession pattern.

---

Note

Overview
Fixes excess TCP/TLS setup when an app runs multiple LaunchDarkly environments by sharing a single URLSession across all DarklyService instances instead of creating one per environment.

Session configuration (TLS 1.2 minimum, no URL cache, reload-ignoring cache policy) moves from init into a static DarklyService.session; each service still sets per-request headers, timeouts, and delegates on individual URLRequests. Flag polling, event bulk posts, and diagnostic posts to the same host can reuse HTTP/2 connections across environments. Streaming is unchanged—event sources still use their own sessions.

Adds a spec asserting two services with different mobile keys use the same session object (===).

^{Reviewed by Cursor Bugbot for commit e03f090. Bugbot is set up for automated code reviews on this repo. Configure here.}