I am no expert on these devices, but in looking at the D4E1/D8Eq modules that you reference, it looks like you could possibly take the incoming E1 circuit and break it into two fractional-E1 circuits.
Then you could connect one half of the E1 to each router, but you would be permanently halving your throughput on that particular E1. Perhaps you could send 20 channels to the primary router and only 3 to the backup router, instead of splitting the E1 in half exactly.
However, two caveats:
1) I would reach out to RAD and ask them if this is a supported/intended use for their device. It is very possible that the device simply muxes/demuxes the circuits, but provides no supervision in the sense that your CPE is currently doing. That is to say, an E1 from your carrier may not be able to establish connectivity when connected to this device.
2) In my opinion, you're fixing the wrong failure scenario. I've seen the carrier's E1/T1 go down FAR, FAR, more often than a failure of the CPE. (This seemed to be true both when I was the carrier and working in the enterprise sector.) I would look at carrier redundancy first, then look into duplicating your edge equipment.