This is a posting that I wrote on the Canard Aviators Forum in the Summer of 2003, as part of what is a regular series of debates about Auto vs. Aero engines for experimentals:

Here are the key factors that I think we all have to consider when
evaluating our powerplant decisions:

1. Operational Data
I think that a major factor in the debate between Aviation and
automotive powerplant proponents is operational data.
A large amount of operational data exists for aviation powerplants in
aviation use.
A very large amount of data exists for automotive powerplants in
automotive use.
A (relatively) limited amount of data exists for automotive engines in
aviation use.
Given the different operating regimes of aviation powerplants
compared to most road car powerplants, not all operational road use
data is relevant to aviation and vice versa.

2. Vendor/factory support
Liability issues and unfamiliarity will probably prevent Mazda (for
example) from ever properly supporting the use of rotary automotive
engines in aviation. Support is therefore going to end up with the
specialist guys.

3. How Much Do You Want To Be A Pioneering Engineer?
There are engineering challenges that are different for aviation use.
The need for a PSRU is usually one (significant) additional issue.
Another issue is installation packaging. The firewall of experimental
aircraft is often a dimensionally limited CG-sensitive zone (the
latter factor much more so than a road car). For example, while it is
possible to install a Mazda rotary in a Long-EZ (Ron Gowan has one),
I am betting that the packaging is really tight. Gary Spencer has a
neat Ford V8 installation in his Long-EZ, and that is a very fast
plane. However, his installation is also very tightly
packaged. IMHO installing a road car powerplant in a tandem canard
pusher is akin to designing an installation for a single-seater
racing car. For the Cozy and Velocity it is less of a packaging
issue, and more of a general engineering issue of balancing the
installation and matching solution components. However, it is still
experimental engineering as long as we do not have standard
installation configurations, components and packaging.

I am contemplating whether to install an automotive engine in my
future canard pusher (the one I have not started yet, since I'm still
learning about them by flying one). Here are some things that I am
thinking about, and issues that I need to feel more comfortable with:

1. Calculate, don't guess
I was impressed by reading about Greg Richter's Cozy Mazda, mainly
because Greg, being an engineer by training, did the math about the
cooling system instead of guesstimating radiator and cooler sizes. If
you're going to be an engineer, you need to work through stuff.
Guesstimating is not engineering...

2. Test the plane with an aviation engine, then switch to the final
powerplant
I would like to see some sort of universal mounting pattern created to
support installation of both aviation and automotive powerplants
without requiring modification of engine mounting points, or some way
devised of installing mounting points for both types of powerplant on
the firewall.
From the perspective of reducing the number of variables, that allows
for better engineering of ground test and flight test programs. You
flight test the plane with an aviation powerplant (preferably one
that worked well on another aircraft), then switch to an automotive
engine when the plane is debugged.

3. Builder funding of test programs
One thing that I think the experimental community needs to seriously
consider is collaborative funding of test programs. If a bunch of
people could fund a ground test cell for a standard (say) rotary
engine package, and have an aviation rotary expert run a
comprehensive long-duration test program based on a variety of duty
cycles and simulated operational conditions, we could accumulate a
large amount of test data rapidly, and also identify component or
engineering weaknesses in engine installation solutions.
At the moment nearly all of the testing of engines and installations
is occurring "for real", usually above the ground. To use a software
development analogy; that is like testing all of your code changes in
Production (which leads to the old joke that Production is the best
Test environment you have - it is, but only if you don't mind
explaining the failures to the customer...). It also reminds me of
Linda Ronstadt's comment (after one of her LPs went platinum) that
she had finally learned how to sing; it was just too bad that she had
to do all of her practising in public (referring to her previous less
successful LPs).
The number of non-aviation engine units installed in experimental
aircraft in the USA, and the number of different variations, puts us
into the specialist motor racing zone in terms of engineering,
testing and operations. Most successful racing teams working with
custom components spend a significant percentage of their budget and
time testing components before installing them in cars. They also
spend money on testing before racing to ensure that the whole package
works, but they go testing with the expectation that most of the
basic car components are going to work. It's difficult to debug a
race car if the engine keeps failing... Bottom line is that they do
not want surprises in "for real" operating conditions. Neither
do I. "The Sound of Silence" is my least favourite song when applied
to aviation.

Is funding collaborative test programs moving us away from the true
experimental ethos? Well, maybe it is. However, I see flying as an
endeavor involving the assumption of calculated risks, not a daily
trip into the unknown. (I leave it to others to determine if that
makes me a true aviator or a wuss).
We all make decisions about risks every time we fly. The better our
decision inputs (data, experiences etc.) are, the higher the quality
of our decisions ought to be (yes, I know that the NTSB database
shows that some folks are bad at analyzing any sort of input...).
One thing about RAF and Scaled that continues to impress me is that
they are always prepared to collect data, analyze it, continually
challenge technology and solutions, and be very conservative. In view
of the safety records of RAF and Scaled, that is an approach we would
do well to emulate when evaluating, building, installing, testing and
operating powerplants.