Flashing red light=bicycle, potentially quite close.
Solid red light could be a moped a mile or two up the road
I am certainly convinced by the argument that in drivers' minds flashing-light=bicycle [haven't seen any research though ... ]
The trouble here is that you are focusing on the scenario of being overtaken on an otherwise empty road. that's not how most accidents happen. most accidents are at junctions, and are usually due to a combination of factors.
e.g. (real anecdata alert:) the only rear end collision I have been near was due to a vehicle overtaking a slower car, not realising there was a bike ahead (or so the driver claimed).
Drivers are usually evaluating a number of data inputs. They may not look at the cyclist they are going to hit for very long. there may be other red lights in their vision.
So here's my conclusion: the evidence exists that flashing lights make motion prediction harder (I didn't make this up). What hasn't been proven is whether this causes more risk to cyclists than the (possible) benefits of flashing lights signalling the presence of a cyclist.
Currently my opinion is that best practice may depend on conditions. in daylight flashers on bikes are probably a good thing. If you're riding on unlit rural roads with NO other cyclists - yup, possibly good, due to more drivers noticing you.
Riding in a group? Definite no-no.
Riding in urban areas, with other cyclists about? Probably a bad thing.
I'm sure we could concoct a number of other grey areas.
An experiment you could do is to spend 3 hours at a (nightime) Audax control, then try overtaking the field again; at some point you may encounter a group of riders with flashing lights on an unlit section. Tell me if their numbers and positions would be easier to assess with steady lights.
(I still can't believe anyone thinks the irregular flashers are a good idea.)