perf regression from adt_significant_drop_fix

[pnkfelix]

Spawned off of: #90218 (comment)

• regression-31157 check regressed by 4.7% on incr and 5.3 on full.
• issue-46449 check regressed by 1.89% on incr-full.
• wg-grammar regressed by 1.3-1.4% in a bunch of scenarios.

Its possible that this is inherent in the change injected by PR #90218. But its also possible that we could fix things to remove these regressions.

[Mark-Simulacrum]

Closing as fixed by #90845, though we may end up reverting it.