I am not familiar with this technology, but from the limited amount I know, it probably is worth pursuing. Having used a stereoscope with aerial photos for many years, there is substantial improvement in 3d viewing and ability to detect heights from of stereo pairs over single photos. We have used LiDAR for many years and enjoyed its improvements in detecting ground surface heights, and I have not heard if anyone has somehow used LiDAR pairs to produce the parallex as LiDAR stereo photos. LIDAR pulses are absorbed by water, so not useful for gravel bed morphology unless channel was dry. Green LiDAR has been used to detect stream channel details under water as long as stream has clarity, but again the one snapshot analysis allows for some parallex and analysis issues if not vertically over the stream. I am not sure the extent of green LiDAR use for vertical change detection in stream channels. In the stream geomorphology courses developed by Dave Rosgen, we learned some techniques to install channel chains that can be used to measure extent of vertical aggradation or degradation at georeferenced points. Rivermorph was the software used to store this type and other channel measurement data. To some extent, channels during floods can vertically degrade and aggrade within a single event, and may continue to adjust somewhat even as water level declines.
To test your ideas, and not have to wait for the flood to come, I would probably choose a braided stream, as they adjust more frequently due to the sediment loading, and maybe you can find a braided system with gravel substrate. In Rosgens stream classification system, I believe that would be a D4 channel type (Braided with gravel as dominate substrate). Another approach might be where they are conducting instream gravel mining or dredging, as these channels are adjusting up and downstream to this activity.
I think the simple answer is yes, you can use DInSar to detect the vertical displacement of gravel-bed rivers as a result of flood events. However, the more complex answer is that it depends on the size of river, the degree accuracy you require with regards to the vertical displacement of the gravel and whether you want to use freely available remote sensing data or earth observation (EO) data which have been undergone significant post processing, which are likely to be more accurate than freely available EO data, but will also be relatively expensive. We have looked at some of these issues in a project we are undertaking to monitor tailings dams in Peru using remote sensing data called DAMSAT see: https://damsat.org/
This paper which came out of the DAMSAT project may be of interest Article The potential to reduce the risks posed by tailings dams usi...
We also have a preprint paper available called "DAMSAT: An eye in the sky for monitoring tailings dams" which is about to be published in a special issue of the Mine Water and the Environment Journal on tailings dams. I can send you a copy of this paper if it is of interest.