Thanks to all the people that helped.
This is the way I decided to write it. Not the most elegant way but runs really fast; .04 seconds to update 200 records. I am building a string of keys. mk_ivinvt holds "AAAAA","BBBBB", etc. up to 120 keys at a time.

My question now is: Why the heck does

IN (SELECT K_IVINVT FROM C_CHNG)

take the same 12-15 seconds as the original query.

Where using

in (<<mk_ivinvt>>)

take only .04 seconds???? It seems they should both be the same but then, as we all know, software works in strange and mysterious ways.

set enginebehavior 90
mseconds = seconds()
*
mrecno = 1
use c_chng
do while not eof('c_chng')
  mk_ivinvt = ''
  mcnt = 1
  select c_chng
  goto mrecno
  scan while mcnt <= 120
    if empty(mk_ivinvt)
      mk_ivinvt = mk_ivinvt + '"' + k_ivinvt + '"'
    else
      mk_ivinvt = mk_ivinvt + ',"' + k_ivinvt + '"'
    endif
    mcnt = mcnt + 1
  endscan
  mrecno = recno()
  *
  TEXT TO mSqlCommand noshow pretext 15 TEXTMERGE
    update ivinvt
    set allocated = allocated + c_chng.chng
    from ivinvt, c_chng
    where c_chng.k_ivinvt = ivinvt.k_ivinvt
    and ivinvt.k_ivinvt in (<<mk_ivinvt>>)
  ENDTEXT
  &mSqlCommand
enddo
*
wait window nowait transform(seconds() - mseconds, '99.999')