Here is the entry for the site from Rivet and Smith's THE PLACE-NAMES OF ROMAN BRITAIN:
- Ravenna, 10630 : ARGISTILLUM.
DERIVATION. Williams finds the origin of this name in a British word which has given Welsh gwystl (Irish giall) 'hostage', drawing attention to Holder I, 1993 (*geislos, *geistlos), and the modern Welsh place-name Arwystli; there is prefixed *are- *ar- ' in front of ' (as in Arnemetia - see AQUAE ARNEMETIAE - and Ariconium). The meaning is thus perhaps 'at the hostage', 'with a folklore reference now lost'. This seems possible, and we have nothing better to offer, even though this makes difficult sense. It should not be forgotten that in a gravely miscopied text like Ravenna there are no certainties when this text alone gives us a name. The present entry could be a garbled version of Ariconium, on the ground that this stands next to Gloucester in Iter XIII of AI, just as in Ravenna Argistillum stands next to Gloucester at 106,29. It is also possible that a root in Celtic *arganto- 'silver', which is common in place-names, is involved.
IDENTIFICATION. Unknown, probably near Gloucester.
Arwystli, while it looks attractive, is scarcely possible, as this Welsh cantref is quite a ways from Gloucester (see map below).
The other suggestions are not very good, either.
If we can allow in this one instance for a sort of hybrid Latin-British name, could this not be for *Ar-castellum? If in full Latin, it would read 'juxta castellum', i.e. 'near, close to, near by, hard by, by the side of' the castle. Castellum here would, of course, designate a Roman fort or fortified settlement. Castellum is attested earliest in Welsh with spellings such as cestill, kestyll, gestyll.
My guess for such an Ar-castellum in the vicinity of Gloucester would be the Kingsholm vexillation fortress:
Kingsholm North of Gloucester, the Site of a Later Saxon Palace