I have occasionally wondered about something that resembles a PEZ dispenser for chip parts....
a simple plate, just slightly thinner than the component, arrangement with a hole in it (fancy shapes allowed) sliding perpendicular to a tube of parts would be one way - a rubber bung (which is only for example, rubber has too much stiction) in the bottom of the tube and low pressure air keeping everything under slight compression would effectively "feed" the parts upwards. Friction from one part to the next allows the (fancy shaped) hole to guarantee that one edge (possibly two if everything is at 45 degrees) are in a known position. A lid over the top of the tube stops them all flying upwards so just a slot either side of the lid allows the plate to move in and out.
and it might be that there's a way of picking them straight out of the top of the tube - or rather feeding them upwards so there is always one at the top once one has been picked
Suppose you started 'tubing up' with resistors. If the tape were laid upside down on a table and arranged to go into a tunnel so that you can use the cover tape as a means to unwind the reel, you could arrange the (slightly bigger profile than the component) tube under a hole in the table so that when a component is in place a simple punch pushes the part down. Here, gravity is your friend and there is no where else for the component to go. With a rubber bung (again, rubber for example) in the tube first, the punch only has to overcome the resistance of the rubber bung so it wouldn't be a punch, more of a push. The punch doesn't even need to be the same size as the part; perhaps something more 'spikey' or 'knifey' so there are no paper / cardboard shavings clogging things up.
You could put as many components in the tube as you like. 0805 is about 0.5mm thick so a complete reel of 3,000 would be 1.5M long (too long) but 1/2 a reel would be ok at 750mm (goes under a table anyway).
It ought to be possible to do that really quickly - 10 per second? (de-bandoleers are that order so I would guess faster) That would be 36000 per hour so it's getting there.
The point being that one 0805 resistor 'jig' could replace many 'feeders' and lower the floor plan of the pick and place machine.
An 0603 could be the same 'jig' with a different hole in the table - move everything over a bit or just a different die.
SOT23s are arranged with the legs pointing away from the direction of travel so they could be done similarly as one would slide onto the top of the previous one.
I can see TQFPs would be a challenge unless they can be picked right out of the top of the tube - and I can see that teeny parts (like 01005) would require a more specialised dooings but they are really special cases anyway. Gaussian theory - 80:20
To put it another way, reels of components have poor packing density when it comes to how many are on a board vs the machine floor plan they take up (as the video mike added shows).