The following citations don't speak to starting slow, but they confirm that speed and accuracy increase with practise.
It is interesting that with practice, [...] limb trajectories get faster and more asymmetric (Elliott, Chua, Pollock, & Lyons, 1995; Khan & Franks, 2000).
Elliott, D., Hansen, S., Grierson, L.E.M., Lyons, J., Bennett, S.J., Hayes, S.J. 2010. Goal-directed aiming: two components but multiple processes. Psychological Bulletin, 136, 1023-1044. PDF page 5.
Decrease in endpoint variability occurs in conjunction with an increase in the peak velocity of the limb (e.g., Elliott et al., 1995).
(Elliott et al., 2010, PDF page 10)
Edit: Found some more relevant sources.
The group exposed to the slowest speed training program was the only group [...] which had significantly lower time-on-target scores from the control group who trained at the criterion speed.
Barbara E. Jensen (1975) Pretask Speed Training and Movement Complexity as Factors in Rotary Pursuit Skill Acquisition, Research Quarterly. American Alliance for Health, Physical Education and Recreation, 46:1, 1-11, DOI: 10.1080/10671315.1975.10615298
The effect of learning a novel skill at specific speeds on performance over a range of speeds was examined on the pursuit rotor. Three groups of subjects were given three days of training: Group 1 at 30 rpm, Group 2 at 60 rpm, and Group 3 at 30–45.60 rpm. Group 4, a control, practiced on a pegboard task during this period. On Days 4 and 5, all four groups were tested for transfer at 30, 45, and 60 rpm. For the most part, Group 3 appeared to perform equal to or better across the range of speeds than any of the other groups.
Siegel, D., & Davis, C. (1980). Transfer Effects of Learning at Specific Speeds on Performance over a Range of Speeds. Perceptual and Motor Skills, 50(1), 83–89. https://doi.org/10.2466/pms.19220.127.116.11
Speed and accuracy are two sides of the same skill coin. although there is a strong and widespread belief in industry that the instructor should focus on developing accuracy because speed will follow, that is not true. The research is clear: if you want speed and accuracy, you must train for both and preferably right from the start of training.
James, R. (1995). Chapter 8. In The techniques of instruction (p. 55). Aldershot, Hampshire, England: Gower.