The epitranscriptome has been mapped in extraordinary detail, and metabolic-labeling assays have begun to measure how fast modifications turn over at the population level - yet the per-molecule residence time of a mark, how long it remains on an individual RNA once written, is rarely recovered. I argue that this variable deserves a name and a place beside the ones we already quantify. I propose to treat the persistence of a mark as an independent clock - mark dwell time, the average time a modification remains on an RNA molecule - running in parallel to the clock of RNA decay. The two clocks are logically distinct, and their relationship, not either alone, determines whether a modification behaves as a static "birthmark" written once for the life of the transcript or a "dynamic switch" written and erased within it. The central claim is that steady-state stoichiometry is degenerate with respect to this distinction: a site reported as "30% methylated" is equally consistent with 30% of molecules permanently marked and with every molecule marked 30% of the time, and no measurement of level alone can separate them. Arranging the major marks - A-to-I editing, m6A, pseudouridine, m5C, and ac4C - in relation to a birthmark-to-switch spectrum exposes how much of their assumed dynamics rests on inference rather than measurement. I close with a constructive roadmap: crossing metabolic RNA pulse-chase with modification-specific, ideally single-molecule, detection to read a mark's level on surviving RNA of known age, and so constrain dwell time directly.
