Correct. Duty cycle has everything to do with it. What might be fine with light amp clipping on 3hrs of sporadic kick drum or HT playback, will probably not be with bass test tones which some of the EDM and other similar genres are pretty much using in some cases.
The 400w recommendation is not the B&C rating. That's from the Powersoft Ipal module literature. The list it as the default setting and somewhere else in their literature for it there is a 700w max recommendation for the 21. That rating is way different from AES testing or pretty much any other that companies are using to rate drivers. That is what we would call a WCS rating. Powersoft offers an actual power limiter that is based on the real average power output of the amp in addition to voltage and current limiters. Most other units do not have this type of limiter.
Most of us probably know this stuff but it bears repeating for anyone reading along. A sine wave has a 3dB crest factor, but music often has a crest factor over time of 12dB or more. If the average power of the sine wave is limited to 400w and the 12dB crest factor music signal is limited to the same average power, the peak output requirement from the amp is 9dB higher for the music signal vs the sine. While you may be clipping a 3200w amplifier on the signal peaks the true average power to the driver could be under 400w at the same time. In fact most of the time this is how things operate and how drivers survive connected to amps capable of thousands of watts of power. It's all about how much time.
The 23 second long measurement sweeps I use do occasionally kill a driver due to excursion or a burnt coil. A 23sec sine wave at the maximum sweep level I use is very tough but at the end of the day it's only 23 seconds and most drivers have no problem with it once or twice and are allowed to cool off immediately afterward. If I were to loop that same measurement signal 13 times in a row it would burn up the drivers in under 5 min. Once we are talking about hours of playback at high volume the drivers have a chance to fully heat soak the motors and coils and all it takes a couple of low crest factor or higher than average level signals to toast the already hot coils.
With all that said I've found most of the modern pro drivers to be quite resilient and durable including the 21DS115, so I'd have to assume the OP's driver was taking quite the beating when it burnt up.