Release LoadGen Api, Appliance and MCP Server: 1.0.0.12207

LoadGen General

• No general changes in this release.

LoadGen Api

• New — LoadGen Score and a built-in management report. Every completed run now gets a LoadGen Score (0–100) summarizing reliability, responsiveness (Apdex), consistency, trend versus a baseline, and infrastructure headroom, computed server-side and shown as a chip on the Results grid. A polished "LoadGen Load Test Report" now ships with the product (no setup): a cover page, an executive summary with automatically derived key findings, the score with a per-component "why this score" breakdown, charts (measurement timelines, logon phases, resource utilisation, score-over-time), per-application detail with success/failure pies, and a full measurement appendix. A new measurement catalog (friendly names, units and targets per datasource) means reports and analytics show meaningful names and units instead of raw identifiers, and you can tune them.
• Analyse — Trends and Percentiles reliability. The Trends view no longer sits on a "Processing…" spinner or silently returns nothing: it aggregates each run server-side (one value per run instead of streaming every sample), explains an empty result, accepts an explicit date range, and no longer misses measurement-only runs. The Percentiles view now flags when a measurement's values are timeout-capped (bunched at a ceiling while failing), so a capped result isn't mistaken for a fast, consistent one.
• Uptime alerting — end the alert storm, add recovery notices. Uptime alert profiles can now notify only on state changes instead of on every check throughout an outage, send a recovery notification (with the downtime duration) when the condition clears, and re-notify on a chosen interval while still down. New recovery hysteresis requires several consecutive successes before a check is marked healthy again, so a flapping service no longer opens and closes incidents repeatedly.
• Alerting — webhook templates and provider presets. Webhook alerts now honor a custom body template (with token substitution) and ship ready-made presets for Slack, Teams, Discord, PagerDuty and Opsgenie, including the variables stateful providers need (so a recovery resolves exactly the incident its alert opened).
• Uptime — Heartbeat (dead-man's-switch) checks. A new Heartbeat check type watches jobs that silently stop running: an external job (cron, backup, scheduled task) pings a unique token URL, and silence beyond its interval plus a grace window opens an incident. Includes an explicit "fail" ping and token rotation.
• [#9320] Added multi-profile tests: a single test run can now carry up to 3 load profiles running simultaneously from the start, each with its own scenario, so independent workloads run together in one execution (the one-run-at-a-time rule is retained by design; the profiles must not share an agent). Schedules can carry companion profiles too, and the Results page now shows whether a run was started manually or by a schedule.
• SessionSight — reliability and shareable sessions. Fixed a SessionSight analytics rollup that could get stuck retrying and stall the pipeline; it now processes history in bounded chunks with back-off. Added shareable session links — a team link or a time-limited public link, cryptographically signed — with optional email sharing.
• [#9316] Fixed a VDI Agent getting permanently stuck on "Unavailable" after an initial misconfiguration, which previously needed a full appliance restart to clear. A failed status probe now re-dials the agent and reports an accurate Offline within a few seconds; once the agent is fixed it comes back Online on its own.
• [#9315] Fixed the VDI Type selection on the Add/Edit VDI Agent dialog not being saved — it now persists and reloads correctly.
• [#9308] Fixed the per-user Agent assignment in a load profile not persisting (it reopened as "-- auto --" after saving). You can now pin a specific agent to a user — with a fallback choice of overflowing to another agent or failing just that user — and the assignment is saved and honored, with any failed-to-pin users surfaced in the live test cockpit.
• [#9313] When a test is already running, Run Test now reports the blocking run by its load profile name and points to the Active Test page, instead of showing a raw execution id; the Active Test view also shows when a run was started by a schedule.
• [#9311] Renaming a Windows Run-As credential profile now updates its link on every agent that uses it; previously the rename silently broke Run-As logon at the next test start.
• [#9310] Deleting a Windows Run-As credential profile that is still linked to one or more agents is now refused with a message naming those agents, instead of silently stranding their Run-As reference.
• [#9271], [#9317] Platform Tokens (Credential Profiles): platform-issued credential mirrors are now removed when their owning appliance or enrollment token is deleted, and orphaned legacy entries are cleaned up automatically, so the tab no longer shows dead entries with a broken "Manage" link.
• [#9309] A Run-As logon verification failure now reports an actionable reason and points to the agent log, instead of a bare "Logon failed (no additional detail from agent)."
• [#9314] Corrected the VDI Agent management port shown in the Add VDI Agent dialog and the API documentation to 4840 (a VDI Agent is managed on the same port as a Full Agent).
• [#9305] Fixed generated report exports potentially carrying an evaluation watermark.
• Appliance boot no longer hangs after saving cloud licensing credentials. On a network that blocks outbound access, an appliance could hang on start with "Unable to connect to the API. Please check your connection." once cloud licensing credentials were saved. Startup now bounds the cloud licensing refresh and continues booting (retrying in the background), and can use the OS-configured corporate proxy for outbound licensing calls. Ensure the appliance can reach LoadGen's licensing cloud (api.loadgen.cloud and loadgencloud.b2clogin.com) for activation.

LoadGen Appliance

• [#9319] Fixed VMware Horizon and AVD (WVD) load tests hanging indefinitely (stuck at "ramping up… Agents (0 slots)"): the appliance now publishes the inbound port (4841) those sessions dial back on, and a new watchdog fails such a run fast with an actionable message when the session hosts genuinely can't reach the appliance, instead of hanging forever. Existing appliances pick up the port mapping on upgrade.

LoadGen Launcher Appliance

• [#9302] A Windows Server guest that came up as an evaluation edition (which cannot be activated) is now automatically converted to the licensed edition before KMS activation, instead of leaving the activation panel dead until an admin converts it by hand.
• [#9294] Added full log downloads from the Launcher Appliance detail page — the complete Core Agent log, and a zip of all Windows guest logs — for support, alongside the existing bounded log tails.
• [#9292] A guest reboot or hostname change is now blocked with a clear conflict message while a Windows guest job — such as a Full Agent install — is in progress, so it can no longer tear down the operation mid-flight.
• [#9321] The Launcher Appliances list now detects a powered-off appliance and shows it Offline immediately (including on a manual Refresh), instead of continuing to show it as active.

LoadGen MCP Server

• [#9320] The start_test tool now accepts additional load profiles, matching the new multi-profile test capability.