The industry's standard term for this phenomenon is "glitch", introduced into the literature in Cooper & Krishnamurthi's 2006 paper on their reactive system, FrTime. They define a glitch as

where a signal is recomputed before all of its subordinate signals are up-to-date

The major review paper Bainomugisha et al, 2012 A Survey on Reactive Programming catalogues 15 reactive systems of which 4 are found to be glitchy. All of the sound libraries feature some form of "pull" based workflow.

The "push vs pull" distinction is mostly valid in that pure push-based reactive systems are likely to glitch. In practice the modern commodity signals algorithms such as preact-signals use a mixed "push + pull" strategy to optimise traversal of the invalidated graph whilst preserving glitch freedom - these strategies are explained by Ryan Carniato and Reactively.

RxJS is a now relatively rare example of an irredeemably glitchy library as explained in this answer - how to avoid glitches in Rx - it's puzzling what use cases it might be aimed at.

There's a standard posting on this subject here: Terminology: What is a "glitch" in Functional Reactive Programming / RX?