The tenon can be dimensionally perfect and the joint will still fail. That's the part nobody mentions when they hand you a list of ratios and tell you to start cutting.
The problem is that fit is not a dimension. It's a condition. Dimensions describe the geometry. Condition describes what's actually in contact when the joint comes together under load. They're not the same thing.
A tenon has six surfaces, and they don't all do the same job.
The cheeks — the two wide faces — are your glue surface. Long grain to long grain across the full cheek area. This is the best glue joint wood offers. It will outlast the surrounding wood if the fit is right. Right means even contact across the full face, no rocking, no gaps. The test isn't a ruler. It's a firm press by hand and a look at the compression line when the joint seats.
The edges, the narrow faces, don't bear load. They center the tenon in the mortise. Small clearance is correct. They shouldn't rattle, but they don't need to be snug.
The shoulders — all four of them — resist twist and racking. When the shoulders bear flat against the mating surface, the joint can't rotate. When even one shoulder is floating, the joint pivots around the ones that do make contact, and you've got a racking problem waiting for the right load.
The end gets 1/16" clearance to the mortise floor. It should never bottom before the shoulders land.
You can cut a tenon to exactly the right thickness, length, and width, and still have cheeks that only contact in the middle and two shoulders floating in air. That tenon will fail. The dimensions were fine. The condition was not.
The shoulder problem is the one that bites most often.
Table saw shoulders are close to square. Close. The wood isn't a machine. One shoulder comes out a few thousandths proud of the other three, the joint bottoms on it before the others can engage, and the joint looks fully seated. It isn't.
The test: press the joint home by hand, then wiggle it — up, down, left, right. No rocking means all four shoulders are bearing. Any rocking means one isn't. Find the proud shoulder and pare it.
One important thing to get right here: fix the tenon, not the mortise face. Relieving the leg face to accept the proud shoulder shortens the tenon's working depth. That's the wrong direction. The tenon gets pared.
Species changes the tolerance, not the standard.
In hard maple or white oak, the mortise walls don't compress at assembly. A tenon that slides home under firm hand pressure is correct. If it needs a mallet, you're forcing the walls outward before the piece has seen a day of use.
In cherry or walnut, the walls compress slightly under clamping. A fit that reads snug in the dry assembly will typically close to the right contact density with glue and pressure. A bit more tolerance going in is acceptable.
In softwoods, pine, cedar, WRC, the walls compress readily. What works in hard maple would crush the walls in pine. Aim for what would feel slightly loose by hardwood standards. The clamp does the rest.
What doesn't change across any of those species: all four shoulders, full contact, no rocking. That's a geometry requirement, not a preference.
On loose tenon systems. A Domino or Beadlock makes the same mechanical joint. Same glue surface area. Same shoulder geometry. The tenon is factory-dimensioned, which eliminates cheek thickness variation and shoulder squareness as error modes. It's a different error profile, not a lower standard.
What it doesn't solve is shoulder bearing. If the mortise slot is cut a hair out of position, a shoulder floats. The verification process is identical to hand-cut work. Press it together, wiggle it, check the shoulders.
Measure the dimensions. Then check the condition.
Cheeks in even contact. Edges with small clearance. All four shoulders bearing flat. End 1/16" short of the mortise floor. If the joint rocks, find the proud shoulder and pare it before the glue comes out.
That's the whole test.