The NFC used by Twitter goes through Canonical Decomposition followed by Canonical Composition¹. This implys to me that decomposition splits it and then it is inelligible for canonical composition. The best way to track what it does is the Unicode Character Database (UCD). Searching the UCD I can see U+FB31 is in the Composition Exclusions file under the Script Specifics heading. That heading² is described as:
canonically decomposable characters that are generally not the preferred form for particular scripts.