"The basement always comes up to meet the ceiling"
To further explain this statement, let's expand the realm of possibilities here.
Let's take our typical Subbie turbo owner who, while having the same 85lb relief HV oil pump, sees 100+ startup cold pressure. I'll venture to say that any cold start produces this starting pressure.
Here the ceiling is raised above the 85lb relief limit. This occurs since the volume that can fit through the relief port is less than Subaru felt sensible at start up visc to deprive the engine of in the maintenance of that 85lb spec.
This will maintain the differential across the filter.
85-23= 62
105-23= 82
The only difference at the exact same volume(s) is a matter of viscosity. Use a 0w-10 oil ..and this would not show the same pressures since the radically reduced visc would fit more easily through the engine and the relief port.
Here's an alternative way of viewing it. It requires a bit of massaging ..and is not offered for "the truth" of what necessarily occurs ..but if you eliminate a few roadblocks it can (perhaps) give you something else to relate to in how it appears.
important side note: I have never studied any of properties of Non-Newtonian vs. Newtonian physics as applied to fluids. I've asked questions of those who do and massaged what they've attempted to demonstrate into agreed to parallels that I can relate to/integrate. That's why I qualify everything with that it is not presented for "the truth" ..but merely to demonstrate the rationale/reasoning applied.
It's almost like a transition from a Non-Newtonian state to Newtonian (for all I know, Newtonian fluids DO transition from a Non-Newtonian "state" to a Newtonian state all the time
) ..except that this is usually demonstrated with a spinning spindle in a beaker. A Newtonian fluid will shear and form a bunch of circles that slow as they move to the outer edge of the beaker. Each layer producing friction to the next ..causing it to start moving. The faster the spindle turns, the mean velocity of the fluid increases.
Now take cake batter and do the same thing. You'll form a very small circle of motion ..and there will be no cascading movement toward the outer edge of the beaker.
Now you just replace the Non-Newtonian fluid in a beaker with a Newtonian fluid that represents enough mass to effectively simulate the same para-effect through a conduit. For a moment, consider it even the lightest fluid you can think of ..in any BIG conduit that you can think of ..but make it a mile long ..just to make 1gpm equating to several thousand pounds of fluid having to be moved at 1gpm between inlet and outlet. I think that you can figure that it's going to take a very long time for any pump to bring that mass up to speed. It may never occur and you'll be in a perpetual relief. Now just tone that down a good bit. You'll reduce the time lag ..or the "phase" relationship between pressure applied ..and volume moved.
Other views:
Think of a steam locomotive spinning its wheels at startup. A muscle car doing a burn out. Do they not parallel maximum applied pressure/force until the motion matches the applied power?
Let's just measure the time that a drag car has less than 100% traction. Isn't there a "differential" that evaporates after full traction is achieved? Now consider viscosity the mass of the car being pushed. Too much mass, longer burn out duration. The longer the "differential" will be maintained.
This is the same with torque converters ..except that they never completely hook up (without a mechanical clutch). You start off with a massive differential that evaporates as the mass gets accelerated.
Boat props ..same deal. It's this "out of phase" relationship that produces the PSID across your filter in most of the elevated PSID events that it will ever see.
Again, non of my examples are stated for "the truth". It's merely to offer other events that may allow you to get a grip on how relief events effect PSID and ..ultimately, filter bypass events.
There are other factors in the mix over the full scope of how a filter reacts in terms of pressure differential. One is loading. This effectively reduces the media ^2" and increases the impact of volume (at visc) through it. It WOULD simulate higher volumes up to the point where volume may be significant enough to make the filter a factor. In my experience with used filters, this still has limited impact ..but YMMV if you're dwelling in higher volume states with extended loading on the filter.