CDO

I was working on removing the ENSO signal on variables and noticed a strange behaviour. Using the attached era_sst.nc file, if I try to remove the ENSO effect within a single piped expression, I get a different result than if I create a file with the ENSO effect and then subtract it.

# Compute a enso-like index
cdo -L -fldmean -sellonlatbox,190,240,-5,5 era_sst.nc enso_temp.nc
cdo -L -ymonsub enso_temp.nc -ymonmean enso_temp.nc ensno.nc

rm enso_temp.nc

# Create random slope and intercept terms
# in reality these would come from linear regression
cdo -L random,era_sst.nc,42 slope.nc
cdo -L random,era_sst.nc,45 intercept.nc

cdo -L enlarge,era_sst.nc enso.nc enso_field.nc

# Expression to compute the effect of ENSO 
enso_effect="-add -mul slope.nc enso_field.nc intercept.nc" 

# Compute residuals by piping the whole expression
cdo -L -sub era_sst.nc $enso_effect residual_pipe.nc

# note the message 
# cdo    sub: Filling up stream2 >(pipe1.7)< by copying the first timestep.
# era_sst.nc and the result of enso_effect should have the same number of 
# timesteps. Why the need to fill up the second stream?

# Compute residuals by saving the effect to a file and then subtracting 
cdo -L $enso_effect enso_effect.nc
cdo -L -sub era_sst.nc enso_effect.nc residual_seq.nc

# The differnece is not zero!
cdo -L info -sub residual_seq.nc residual_pipe.nc | head

Result of the last expression:

    -1 :       Date     Time   Level Gridsize    Miss :     Minimum        Mean     Maximum : Parameter ID
     1 : 1980-01-01 00:00:00       0    10368    3911 :  -0.0039062  2.4286e-05   0.0039062 : -1            
     2 : 1980-02-01 00:00:00       0    10368    3911 :  -0.0033875    0.029031    0.061271 : -1            
     3 : 1980-03-01 00:00:00       0    10368    3911 :    -0.24288    -0.12062   0.0037231 : -1            
     4 : 1980-04-01 00:00:00       0    10368    3911 :    -0.86571    -0.43523   0.0022888 : -1            
     5 : 1980-05-01 00:00:00       0    10368    3911 :    -0.78418    -0.39365   0.0022888 : -1            
     6 : 1980-06-01 00:00:00       0    10368    3911 :    -0.57350    -0.28754   0.0022888 : -1            
     7 : 1980-07-01 00:00:00       0    10368    3911 :  -0.0029907     0.12092     0.24309 : -1            
     8 : 1980-08-01 00:00:00       0    10368    3911 :  -0.0029907     0.38596     0.76908 : -1            
     9 : 1980-09-01 00:00:00       0    10368    3911 :  -0.0029907     0.29365     0.58588 : -1 

Am I misunderstanding something important about how cdo pipes work? Or is this a bug in cdo itself?