If you want to give your brain some serious exercise, take a look at this paper. It explains the basics of the transition spike problem, but in the end it is leaning towards a solution of using a Closed Transition Y-D starter, which is an expensive solution compared tousing a solid state starter now.
This same company used to discuss the cheaper solution of rolling the conductors, and they sold a cheap little device called a "Leading Phase Indicator" to help you do it. It appears that since they have started selling solid state starters they no longer support providing a way to help mitigate the transition spike on Open transition Y-D any more.
Rolling means taking the three leads and shifting them over one terminal, yet keeping the same rotational relationship. For example, move A to B, B to C, and C to A. It will still rotate A-B-C. Here is a link to Franklin Electric's instruction sheet on doing this. They promote it as a way to try to match natural motor winding unbalances to voltage unbalances in order to have a more balanced current, but the principal is the same.
Rolling changes the relationship to the inherent phase shift that happens in Y-D transition. So if your voltage was lagging the current, which may have been contributing to your spike, you can make it lead the current during transition. You still get a spike, and if you read that paper it will tell you that it can be as high as roughly 3x the LRC, but you may minimize it to a lower value. Unfortunately they stopped publishing the older paper that gave the calculations for the lower value, but it was better.
Note that this is not 3x FLC, it is 3x LRC, which is usually already 6x FLC! So that means that you can have a transition spike of up to 1800% FLC when switching from Y to Delta. That can give the contacts (and motor windings for that matter) a really hard time.
"Venditori de oleum-vipera non vigere excordis populi"