There are two things in SPICE that I wish were never invented by the
authors of the language:
1) .global nodename
2) .option scale = value
These two commands can actually be very useful and convenient
for circuit designers, who tend to work on one buffer at the time,
because they can reduce the amount of typing necessary for writing
netlists and/or models, improving productivity.
However, these commands can cause a lot of problems when two
different designs, having conflicting global nodes and/or scaling
coefficients must be simulated together in the same simulation.
This would usually happen in Signal/Power Integrity (SI/PI) simulations
when the engineer is trying to design a system which obviously
involves more than simulating just one buffer at the time.
In my experience, SPICE models are most often written primarily
for circuit designers, and the SI/PI people are not considered in this
process. When the design of a buffer is done, the models are just
handed over to the SI/PI engineers, and that's when the problems
usually begin. To make things worse, by this time circuit designers
are quite often done with the project and are reluctant to go back to
make modifications to the SPICE models on the request of the SI/PI
So my recommendation is to STAY AWAY FROM these two features
of SPICE to save a lot of trouble down the road.