Your instinct on the kill valve is sound, and the pattern is well-documented. one account I tracked jumped tROAS from 350% to 600% overnight and spend dropped 40% within five days. Same magnitude as your jump from 42% to 100%. Expect a cliff, not a gradual optimization curve. Seen this play out more times than I can count.
the mechanism is almost too literal. tROAS pulls bids down hard once the target sits meaningfully above the achievable ratio. it's not discovering new value the way your rep implied. it's simply refusing to enter auctions it can't win at that ratio anymore. the algorithm isn't creative, it's mechanical.
on the ladder question, best practice runs opposite to what you'd think. Start 20 to 30% below current actual, not above it. Then tighten in 10 to 15% steps per stabilization window. that only works when actual is already close to target though.
You're jumping from 0.42 to 2.0. That's not a ladder problem anymore. That's the campaign telling you the auction pool at that value threshold doesn't exist yet. Kill valve is the right call, not patience.
Migration sequence sounds right too. switch goal first, run Max Conv Value with no target through the accrual window, set tROAS once conv value per cost stabilizes. One gotcha though. Smart bidding wants roughly 30 conversions in 30 days per campaign before it reads signal cleanly. Below that, the goal switch alone tanks performance during relearn no matter how carefully you sequence everything else.
On client accounts I manage north of $200k to $300k+ monthly, i never touch a target until that 30 conversion mark clears per campaign. Jumping the gun there usually causes the learning limited spiral people end up blaming on the target itself.
