Issue when stride != width

I'm trying to colour calibrate a sensor connected to a single board computer and I'm running an issue where I'm not sure what the root cause is.

Context: I'm using an RB3gen2 development board from Qualcomm with an imx577 sensor hooked up to it, with the ISP in the CPU disabled. This gives me 4056x3040 images at 30fps in SRGGB10P format. With gstreamer+libcamera I can software convert that to NV12 and show it on the screen:

When I use the 'yavta' utility to capture raw data it says this:

root@qcs6490-rb3gen2-core-kit:/media/nfs# yavta -B capture-mplane -c -I -n 5 -f SRGGB10P -s 4056x3040 --stride 4056 -F /dev/video0
Device /dev/video0 opened.
Device `Qualcomm Camera Subsystem' on `platform:acb3000.camss' (driver 'qcom-camss') supports video, capture, with mplanes.
Video format set: SRGGB10P (41415270) 4056x3040 field none, 1 planes: 
 * Stride 5072, buffer size 15418880
Video format: SRGGB10P (41415270) 4056x3040 field none, 1 planes: 
 * Stride 5072, buffer size 15418880

Note the 'stride' there. Running makedng -p SRGGB10P -w 4056 -h 3040 -m RB3gen2,imx577 frame-000001.bin frame-000001.dng gives me a DNG that Rawdigger shows as such:

Doing makedng -p SRGGB10P -w 5072 -h 2400 -m RB3gen2,imx577 frame-000002.bin frame-000002.dng instead gives me

Going higher than '2400' will trigger a segfault in makedng, even if the filesize is large enough to have stride*height pixels.