Before the "upgrade" I measured the bandwidth as almost exactly 50 MHz with a sharp rolloff above 50 MHz, consistent with bandwidth limiting in software.
Potentially, that could be a concern. I.e., software BWL does nothing to eliminate aliasing, because the damage has already been done by then. (Higher frequency components have already folded back into the pass-band.) That's why there is a programmable LPF in the gain-stage of the front end, in the first place.
However, this isn't unexpected, since the chip they're using has no 50 MHz LPF setting. I'd guess they're using the programmed 70 MHz setting in hardware, which in practice should be fine. But for those paying attention to small details, the 70 MHz models are only down 3dB at ~90 MHz. And therefore the more "conservative" 50 MHz BW (with respect to a 4-channel 250 MHz sample rate), doesn't buy you any more immunity from aliasing, because it's imposed in software.
Assuming of course that swperk is correct (and I believe he is).