The general consensus seems to be don't excavate.
The only approach I have seen that meets OH & S requirements and allows some stratigraphic control is the MacIlroy method where the area outside the well is opened first and the well is excavated from the sides thus allowing any stratigraphy to be identified. This was used on the Cyprus project in Oakland.
Otherwise if you want to remove the fill and there is some water in the well you could try a variation on the airlift used in maritime archaeology.
Iain Stuart
[log in to unmask]