I've made a first order simulation of an oil-can delay in DSP, where the delay time is synchronized with a vibrato wobble frequency. Can sound really interesting. You start thinking about the angular displacement of the read and write heads, multiple heads, frequency response, distortion, etc. but as a real one of these (or Morley/Fender equivalent) was fetching upwards of $750 on flea-bay last time I looked, I may never have a real one in hand to measure. Not to mention the unobtanium based oil!
There's a demo of a Fender model on YouTube where the sound really gets pretty weird when the speed is set at its lowest setting. I think what happens is that the oil film which retains the signal pattern physically on its trip from the write head to the read head starts to drip, causing a random phase distortion "or something".
Note that I like to summarize super technical conjecture with the catch-all phrase "or something" as it allows me to sound erudite and at the same time I can be completely wrong! 
i have one waiting to get rebuilt on my bench, complete with the unobtanium oil. they are really weird!
the modulation is completely unpredictable, they basically run on rubber bands.
the heads are actually rubber, they barely touch an anodized aluminum disc. the oil is what actually transfers the signal. its bizzarre!
i have to take it completely apart to clean it, which is a major drag <standel reverberola or something like that> and i gotta recap it, but there's virtually nothing in there. couple passives, couple pots, couple tubes.
these things really have a sound of their own. a bud plays with hilton valentine <the animals> and has been trying to get it off me since ben lyman hooked me up with it.
thanks sir mark for a most cool posting!!