I feel like it's important enough to mention this to create a new answer even though there's a ton of good information here already, as you mention it's the transient currents that kill ESCs. To be perfectly honest we're well within the rating of even 20A (25A 10 second burst) ESCs in terms of sustained current draw on almost any setup you can imagine. It's not sustained in-flight current that kills stuff. Dynamic unloading means current draw drops pretty quickly as the quad is moving forward even at full throttle, and there aren't too many batteries that can sustain even 100A for more than a few seconds. After 3-4 seconds the battery sags and the current drops away. I've never seen a battery that can sustain 100A for a full 10 seconds, even the highest C rated (for whatever that's worth) packs at high mAh. I've flown some ridiculous setups on even 20A ESCs and drained 1300 mAh from a pack in under 90 seconds.
What kills ESCs is transient spikes. As you mentioned typically those spikes come from blocked props or impacts during rotation that stops the motor. Also, those spikes can come from desync situations between the motor and the ESCs which has basically the same impact as a blocked motor, a current dump into the coils due to stall condition.
The reason I point this out is that a larger torque load on the motors can end up dramatically increasing those current spikes, not on a normal forward flight, but on rapid changes in RPM. If you check out the data on MQTB, you can see the instantaneous spikes on heavier props are dramatically higher than lighter props. Even on fairly light 5" props, current spikes on rapid throttle changes can hit upwards of 150A for a couple of milliseconds during rapid throttle changes. This is under static load, and the spikes can be even higher when the effective angle of attack is even greater (such as rapid changes in direction at full throttle when the quad is moving the opposite direction from thrust.) For instance, I logged on the bench a 2207.5 1800kv motor hitting 170A briefly on 6S during an instantaneous change in the throttle from 11% to 100%.
It's also important to note that the more torque a motor is capable of producing (the larger the motor stator size, tighter air-gap, or stronger magnets, etc) the larger these current spikes can become. Highly loading a larger, torquier motor is going to create larger spikes, given that torque is directly proportional to the current in brushless motors.
Now, this brings us to the solution. If you're using a BLHeli_S or 32 ESC you can decrease the ramp-up power, which largely eliminates these spikes. If you're considering pushing the torque limits of a motor I highly recommend making this change. Even 5" on 6S, reducing ramp-up power dramatically reduces the risk of desync conditions and smoked ESCs, and you're likely to run into similar. Even if you don't instantly smoke the ESC, the transient current spikes increase the day-to-day wear on the electronics, so making this change will likely increase the longevity of your gear in such high torque load situations.