It is most likely a low-voltage low-current glass gas discharge (GGD) tube. It will not conduct at all until a certain voltage is reached, then the gas ionizes and it become effectively a short circuit until the current is removed. This small size is to suppress back-EMF, that is short-term low-current pulses. These devices can have capacitance values as low as 1.5 pf, making them useful in RF and microwave circuits to about 2 GHZ (in an smd package). For higher currents Transzorbs or MOV's would be used.
EDIT 1: Glass gas discharge (GGD) tubes are sensitive to fast rising or falling voltages. Rise and/or fall times less than 10uS can cause the gas in the tube to ionize quickly, at a voltage much lower than expected. It is a great idea to read all the fine details of a GGD before using them. In some cases such as back-EMF the fast rise time is an advantage, as the GGD will clamp it at a lower voltage than its normal slow rise-time clamp voltage.
EDIT 2: This could also be a tiny Transzorb. They are much like back-to-back zener diodes but can only handle brief burst of low currents. Transzorbs are not meant to be voltage regulators.
Well.
Let's suppose you split your 9V wall adapter into +4.5V and -4.5V rails using a virtual ground which we shall name VGND. The other option is to use standard split rails: +4.5V, -4.5V, and a real GND.
Now, we connect all the single supply (SS) chips between +4.5 and GND (or VGND).
All the SS chips' supply current is obviously drawn from +4.5V, and loops back into VGND. This includes your echo module. Therefore,
- your VGND generator should be able to sink enough current to cover then entire supply current of your SS chips. TL074 can't do that, it's an opamp for very very light loads like a 10k resistor.
- VGND is also the voltage reference for the dual supply (DS) chips. Feeding random variable currents into it (from your SS chips' supplies) will inject noise into VGND. It would do the same with a real GND, but the impedance of a ground plane is quite a bit lower than the output impedance of a virtual ground. This might make your layout more complicated if you don't want noise can enter your audio signal chain.
Now, the second point isn't set in stone. If everything is referenced to VGND, and its layout is good, then it will work just as well as a normal GND. However you should be very careful not to have two different references (can happen if a part of the circuit is AC-coupled). For example, if one reference is VGND and the other is a voltage divider between supplies, then as VGND wiggles around due to it being used as supply ground, the other will not follow, and the difference will be injected into your signal.
Note: When one of the chips pumps current into VGND, you can say "VGND has noise". But from the point of view of your circuits, VGND is fixed, since it's the reference. It's always 0V, by definition. The "noise" I'm talking about will appear on both supplies instead. And contrary to a normal design, where you could add filters into the supplies to isolate a noisy bit of circuitry from the rest, here it would be more complicated.
Also, if you use a standard virtual ground chip, the last I checked generated huge class-B distortion on the rails when AC current was drawn from VGND. It is a voltage follower opamp after all, with a class-AB output stage, and usually a very low bias.
Good capacitors are cheap. Virtual grounds are a headache. I would AC-couple everything, and use single-supply everywhere. Much simpler.
Best Answer
Looking at the last page of the DATASHEET tells us that the label is printed upside down.
Refer to the pinouts in the datasheet and ignore the label.
Note the location of the cut corner on the label area.