At present, checks running in suggestion mode recheck the entire document whenever the document changes (e.g. during typing).

Without a speedup mechanism such as caching, this architecture is insufficient to support a large number of suggestions running on a long article.

Certain checks already have caching (TextMatchEditCheck, ToneCheck) but most do not.

SLOs for edit suggestion performance are still to be determined, but even without specific quantitative limits, it's clear performance would become absurdly slow if we deployed many edit suggestions with no performance improvement.