The above is overstating the authority for C. In any language in which CE and SE are pronounced the same, there will be at least occasional fluctuations between them.

The earlier Old French spelling was with C, influenced by unrelated words such as intercede and precede (from Latin cedeo 'I go (from); I give way'), but it changed to S in later Old French. The spelling with C was also often used in Mediaeval Latin.

In English however its use was almost entirely confined to the obsolete Scots sense of 'postpone, delay, be delayed'. This was usually spelt superceid, with a few variants such as superseid and supercid. In all the OED quotes, throughout the Middle Ages and to the modern period, the C spelling is extremely rare except in Scotland. One from 1654 can be regarded as a valid variation, but the one from 1807 is simply a mistake: by then spelling was fixed, as it was not in 1654.

Other forms consistently show S in the vast majority of cases. Supersession has three instances of C between 1798 and 1859, presumably as it's a very unfamiliar word that looks remote from the familiar supersede.

The legal term supersedeas or writ of supersedeas always has S. The first instance in English is in Piers Plowman, 1393. The OED gives no C quotes, though it is sporadically misspelt as e.g. supersedæas.

Supersede is isolated because no other English word ends in -sede, from Latin sedeo 'I sit'; but throughout the history of English it has always had S, with C as a minor variant. The wrong spelling might even be more common in casual usage, and occurs quite a bit in print, but for now we can still say the C spelling is wrong.