How can I correct instability in a push-pull transistor amplifier?

In our EE lab, we had to build the following amplifier. It is from this lab manual, page 40.

A week ago, I built it successfully, with 2n3904/3906 and d44vh10/ d45vh10 for the power amplifiers (Q3,Q4). It was driving an 8ohm speaker and all worked fine. I put my circuit in a box and came back this week to demonstrate its operation to the lab instructor but suddenly my circuit just oscillated at a very high frequency and nothing I did could correct it. I have rebuilt it twice now and somehow it is stable until I apply a signal and increase the volume - then it just goes crazy. At one point it worked perfectly fine when $R_L$ was a 10K resistor. Once I connected the speaker, it just went crazy and the transistors started heating up. What could be causing this instability and how can I correct it?

I have noticed the issue is at the output stage - when I remove the op amps and connect the midpoint between the diode pairs to ground, it still goes crazy so I think the problem is within the transistor part.

Component Values:

• R1=R2=5k
• R3=R4=1ohm
• Vsupply = +/- 15V
• Op Amps: ua741 8-dip
• You should add component values and supply voltage numbers in the schematic. Assuming certain values for those then my bet is that your circuit is OK (it is stable with a light load) but your practical setup isn't OK. When you load the output stage, large currents can flow. There will always be series resistances of contacts and wires which can cause unexpected effects. Can your supply handle the currents? Do you have enough supply decoupling? Does your buildup look similar to how it is done in a commercial amplifier? Open a old audio amp and learn from that. Oct 25, 2017 at 19:31
• Crappy design using $R_1$ and $R_2$. You'll learn why in time. But I'm not suggesting that's not your problem. Aside from good advice already about adequate bypassing, I think $IC_3$ is oscillating and you need to think adding some compensation to it.
– jonk
Oct 25, 2017 at 19:50
• You might add a small resistor in series with the load. Say 0.1 or 0.2 Ohms. I think Jonk is right that IC3 may be oscillating when the real load is applied. This small resistor will isolate the op-amp slightly from the load impedance. Oct 26, 2017 at 4:49