For a launch operator or a program office, the only forecast that matters is the one sealed before T-0. Here are the launch calls that were — Starship IFT-2's vehicle loss, Vulcan/Peregrine's payload loss, and New Glenn NG-3's payload loss after a sealed slip campaign — each dated, each checkable.
For a launch operator, an insurer, or a program office, there is exactly one kind of forecast worth anything: one sealed before T-0, on a named vehicle, with a failure class attached. A post-mortem is free; a dated heads-up is not. These are the launch calls in the corpus that meet that bar — sealed before liftoff, graded against what the vehicle actually did, and checkable by anyone. I'll be precise about which were loss calls and which were slip calls, because blurring the two is the fastest way a track record stops being trustworthy.
STARSHIP IFT-2 — a catastrophic-loss call, sealed pre-flight. On 16 November 2023, before the 18 November window, the advisory named a window-specific catastrophic failure — at a moment when the public framing was optimistic, 'lessons learned from IFT-1.' On 18 November both the Super Heavy booster and the Starship upper stage were lost. No named analyst, journalist, or competitor is on record with a window-specific catastrophic-failure call for that window beforehand. Graded HIT. Advisory: [/advisory/LA-001](/advisory/LA-001).
VULCAN CERT-1 / PEREGRINE — a payload-loss call with an unusual split. Sealed 3 January 2024, the call separated the rocket from its cargo: the Vulcan launch vehicle would perform, but the Peregrine lunar payload would not complete its mission. On 8 January 2024 Vulcan flew nominally and Astrobotic's Peregrine lander suffered a propulsion anomaly that ended any chance of a lunar landing. Calling the rocket good *and* the payload lost, separately, is a harder and more useful conjunction than a blanket 'it'll fail.' Advisory: [/advisory/LA-003](/advisory/LA-003).