Sorry, for technical it will be too complicated to do this. However, it doesn't matter because the "skr-arab" code will remain internal, and you'll see "skr" in the actual sites.
In summary we jsut want to make sure that [skr] is an alias for [skr-Arab] by default, just like [zh] is in alias for [cmn], or [en] is an alias for [en-Latn].
Like for all aliases, they are to be treated equivalently, except for the purpose of fallbacks with BCP47 rules, where the extra script code can be used as a guide to select one of several possible languages using the same script. This does not mean we create duplicate data for localization, but that where data is missing, fallbacks will resolve differently for languages that may have different scripts, even if one of them is the current standard or the most frequently used.
- e.g., English is not written only with the Latin script, it may still be written with Deseret, so [en-Dsrt] is still a distinct variant, but if it's missing it will fallback to [en]
- for Serbian, the Cyrillic script is the default so [sr] can be aliased to [sr-Cyrl] (or [sr-ec] only in Wikimedia wikis which violates the BCP47 standard here) and we don't need to have the two datasets for [sr] and [sr-Cyrl] (they should be identical), but we can still have [sr-Latn] which fallbacks to [sr] (and thus, also to [sr-Cyrl] because it is equivalent by its alias), however it can also fallback first to [sh-Latn] which will itself fallback to [hr-Latn]).
We can consider that for fallbacks, preserving the script is preferable to preserving the base language; so the two aliased codes are not completely equivalent: they are only equivalent in translation datasets, but not in applications that will use them with their own fallback chains to decide the order of resolutions.
If no fallback is needed (because a translated resource is found in any one of the two aliased datasets, which are in fact just the same one), using one code or the other does not matter. So in this TWN wiki, nothing needs to be changed except making sure that there's only one dataset open, the other being locked, or "redirected" to the other code.
So I would recommend keeping [skr-Arab] here, and keep [skr] closed/locked, with a possible redict. On the [skr] portal, we can display the shorter code, but as well all variants including [skr-Arab] as the default variant. The short code will be locked, all translations will be made in the unambiguous [skr-Arab], except if another script is used, like [skr-Cyrl].
It should be noted that [skr] was just approved by the languages committee of the WMF, for creation of Wikipedia at least. For now all Wikimedia projects are still in Incubator.
Here on this wiki, it will remain [skr-Arab] (which is already enabled), even if there are 3 other scripts (including an historic Indic abugida, Multani, which was specifically used for that language; plus Devanagari and Gurmukhi, or Latin transliteration sometimes used today.) This won't block the import and creation of the new Wikimedia wikis (using mainly the Arabic script) as they work already in Incubator: the code [skr] can then remain disabled on Translatewiki.net (and because it was already disabled, there's no need to merge an inexixtant [skr] dataset with the existing [skr-Arab] dataset).