The Z80 does not do 16-bit maths pretty quickly.  It is an 8-bit chip and
it does arithmetic on an 8-bit ALU.  It does 8-bit arithmetic operations
in 4 T-states, and 3 of those are the instruction fetch and decode cycle.
It does "inc hl" in 6 T-states and "add hl,bc" in 11 T-states (only 3 of
which are needed to fetch and decode the instruction in each case).  It
takes 5 T-states just to add the displacement of a relative jump to the
program counter.  It takes 19 T-states to do even the quickest operation
on (ix+d) (that's 7 for the important part of the instruction, 4 to decode
the DD byte, 3 to fetch the value of d, and a whopping 5 just to add d to
ix).

Given that in a 16-bit arithmetic operation it is the upper 8 bits of the
result that are formed last, it is hardly suprising that the C, S, H, V
and two unused flags of the F register are set from these bits.

The Z80 is an 8-bit chip.


Ian Collier