The stress in steel doesn't equal the stress in aluminum.
Thermal stress is not the same as stress due to loads.
and the geometry is different, therefore the equations for stress will be different.

take the thermal expansion coefficient,
times delta T
times a dimension in question such as the overall height
this equals the amount the thing will grow.

there's an equation for loading the center of a plate with deflection,
put in the previous number
somewhere there's another equation for bending moment and shear
add bending moment and shear times cross section is the stress

thens there's the allowable stress for aluminum
that's the number you looked up.
multiply that by the temperature conversion factor and compare

I didn't go into much detail of my calculations since it probably bores most people who do not wear pocket protectors or have electrical tape on their glasses.  Here goes: Youngs modulus relates stress to strain for a given material. Since the assumption is that the steel is constrained by the aluminum case, the length the steel expands is zero. So I actually started as you suggest by determining the length it would expand by being heated from 70 to 250 degrees if unconstrained, and then figured out how much force it would take to squish a piece of steel back to the original length. That puts length on both sides of the equation and length drops out as a factor in calculating stress in this unique case.
The resultant equation is simply stress =coeff of thermal expansion x temp diff x youngs modulus.  It always amazes me how much stress can be developed in this scenario. Often the constrained material will bow significantly if the constraining walls do not deform or fail.
So now we have a piece of steel applying 1,000 psi,,,then 10,000 psi...then 25,000 psi against the outer case wall as it heats up but remains constrained by the aluminum. Aluminum loses the arm wrestling match and shears as the steel finally builds up sufficient stress to exceed the shear strength of the cast aluminum. The aluminum will of course fail at the weakest point, usually where there is a defect such as porosity or sand inclusion.

This was just a proverbial "back of the envelope calculation" to see if there was any way to generate even a ballpark number that would explain it. Perhaps it is too simplistic to reduce the filter to a single element.  

The failure mode looks like the way a frozen soda can blows up in a freezer. Of course, since water is one of those odd materials that expands, instead of contracts, on freezing, it gives the same failure mode by freezing that steel would provide on heating. It expands, deforms the can a bit then explodes catastrophically when the force exceeds the shear strength of the aluminum.

At any rate, I was just trying to make sense out of it. It would be hard to believe that high cycle fatigue of 10-20 psi oil pressure would fatigue the casting.  Constrained thermal expansion is the best  idea I could come up with.  I'm open to other explanations and more accurate calculations that would eliminate this theory.

This is an old thread, but this just happened to me. I changed the oil and filter a few weeks ago. I rode it several times since, including a day with about 180 miles. Then yesterday I was starting it up and POW! the cover exploded.

This is my first bike and it was the first time I changed the oil and filter on my own. I followed the directions and everything seemed in order when I put it back together, but operator error may still have had something to do with it.

Anyway, here's the aftermath:

I'm just glad i wasn't riding at the time, or I'd probably have a few square inches of aluminum in my calf.

... didn't quite follow the directions... ...

All righty then.

That's a mistake I won't make again, and like verslagen1 said, maybe someone else won't either. Which reminds me of a quote I once read: "I like learning from my mistakes, but I'd rather learn from yours."

Thanks for the lesson, y'all!