In a collation rule, a reset string (the "abcd" in &abcd < x...) could in principle contain a part which depends on precontext or on a contraction. Precontext is known to not work currently, and contractions in this place are not tested and therefore probably do not work either.
The CE of b could depend on whether it follows an a (precontext):
&u = a|b
The CE of cd could be the result of a contraction:
&q = cd
Therefore, the new CE for x defined by
&abcd < x
should intuitively be a primary step up from
CE(a) CE(u) CE(q)
While not urgent, and not initially scheduled, it would be nice if ICU collation rules supported this. It may be possible to support it with the &abcd < x syntax, or it may be better to use special syntax, just like for expansions where we treat &mn < y as &m < y/n.