Connect the contributions
Predicting an oscillating system can require a physical account, a model with stated idealizations, a mathematical consequence, an algorithm for the approximation and notation whose variables and units remain interpretable. A measurement that defeats the idealization returns to the physical and modeling premises; computing more accurately cannot repair that defect. Conversely, a numerical error need not invalidate the physical law.
For a claim about a particular test, start with the decision the reader owes. Which motion did the instrument actually sense, what did its record preserve, and what does the calculation prove about that motion? One reliable reading above a limit, after allowing for measurement error, can establish an exceedance. To certify that the limit held throughout, a stored peak must bound the physical maximum, including peaks the sampling or instrument response could have missed. A file holding only the final position can lead to the same limit if a supported physical law supplies the missing bound. For an order to derive a mathematical bound, the proof itself can finish the work. Choosing the relevant scales and model detail
These are different grounds for a connected result: mathematical validity, physical adequacy, computational feasibility and intelligibility of expression. Each matters only where that claim is made. A pure proof needs no physical test, and a notation problem need not start with a physical question. Transferring an algebraic rewrite to a program expression depends on what evaluating that expression does: q + q = 2*q remains an identity, while read() + read() need not equal 2*read() when each call obtains a new value.
The Reference explains how the contributions constrain one another (Foundational.Reference:2.2) and what preserves or changes the field (Foundational.Reference:2.3). The connected methods remain independently usable.