I warn that the following comment may amplify confusion because I don't understand how to think about this topic.
The OP comment edit asking for an alternative got me going:
//Edit: thinking more about this, I would throw it back at you - what would you write for converting the Fahrenheit value to Celsius? It’s one thing to say I’m wrong, which is fine, but you offer no alternative.//
Translation manufactures a false precision, i.e.:
Units translation seems to be a slippery slope: to maintain accuracy in translation context imbues the data with precision that may not exist in the original samples.
134F -> 56.67C
So a report with a precision of +/– 0.5 becomes precision +/– 0.005 via translation.
This seems like a hazard to chain of custody in any translated reports.
As to how important this is, this is open to consideration.
Discounting the contrived precision in context converges uncertainty at the scale of the least precise unit:
56.67C rounded to 57C = 134.6F -> 135F -> 57.222C -> 57C
And when further reported...
+/– 0.5
There seems to be no alternative to maintaining precision in context than to imbue the data with false precision; the precision is required to avoid creeping noise.
Consider additional pitfalls in report hygiene, such as idiomatically omitting the full precision in context for a measurement that falls on a unit interval. Consider 134.000F recorded as 134F, then decontextualized to +/– 0.5.
It seems any chain of custody includes an implicit hazard of well-meaning custodians.
When does engineering goe wonky merely by being translated?
This might seem obsessive, but
being open-minded about what it means to "translate" leads to more questions:
Consider precision regarding inter-disciplinary transposition of data where the context shifts from a relatively narrow scope of scientific research, say collection of temperature trends from ice cores, that's translated into very broad scope state policy, say a schedule of global average temperature change. For example consider an emissions policy based on a target of +1.5C over a 50 year window. You never see this policy written in precision corresponding to the basis of data collected— e.g., written as 1.500C— even though the geological basis intervals range from order 1000–1000000 years.
To me a policy target of order 1.5C seems fishy based merely on the precision of its presentation in scientific context. Shouldn't it be order more like 1.5000C?
I'm curious about inherent distortions that may attend any analytic approach to policy as data are moved been domains of reference.
The OP question surprised me; it seems to be a true conundrum, beyond the reach of any simple "alternative" in context.