You'd pretty much start with a local radiometric survey company (there are a few in Perth) using a 50 litre doped sodium iodide cyrstal pack (a bit heavy for a drone) and scintillation counters in a crop duster flying at 40 m ground clearence dead centre down the road at 70 m/sec for the full 1,400 km road length.
Post process the 256 value gamma spectrum, correcting for cosmic, aircraft signature, remove the mean average W.Australia backround signature, run a full rolling NASVD to peak sharpen and look for 32 and 662 kev twin peaks.
Then you'd start with the local area ground search.
Given the lapsed time already there's a good chance the slug in question was either magpie'd on site or has been picked up in a tire and gone off route.
If you read all that and still feel game for more there's some mad bastard willing to tolerate PhD candidates that want to mix geology and big radiometric data, he even reckons (and I quote):
The integration of these different lines of research has the potential to improve greatly our ability to extract geological information from radiometric datasets for exploration purposes.
It’s called Bayesian Search Theory, and is even more interesting when you consider that not finding it at a given location gives you information about where it might or might not be.
I wonder if analog dosimeters might be a better way to narrow down the search area more cheaply and quicker to set-up. Fly or drive a run. Check dosimeter.
Dunno if there's enough energy expected to set it off at the expected speeds travelled. Obviously they don't go "ding ding ding". Might find some spurious beta-radiation sources?
Here's how I would approach it... Based on a given detector geometry you can imagine it has a detection efficiency as a function of distance from the detector. Now you can imagine the source traveling in a straight line such that it passes some minimum some distance from the detector and at some velocity. From that path there is an expected number of counts that the detector would receive on average from that source as a function of minimum distance. Using that function and the background count rate one can estimate how close the source path needs to be from the detector to reliably detect it. From this you can optimize the altitude and velocity of the detector traveling along the road. I hope this brief explanation makes some sense. The hardest part would be to decide what signal to noise ratio is acceptable. For vehicle mounted it would simpler because you can probably just fix the elevation and only focus on the velocity as a variable.
What if we recreate the journey, use the same truck with the same type of containers and see if we can re-create the malfunction, then start the search forward/backwards from there?
Given the parameters to maximize detection and minimize time/resources, how would you direct a drone or something to find the capsule?
How would that change with 2 drones? 4? 8? As many drones as there are flyable Geiger counters in the country?