Normally you'd make the deadtime large enough to match the datasheet's worse-case switching times for the device across full temperature range.
Note that switching speed changes with temperature and load current, so unless you can predict these accurately there's no way to take advantage of any device matching. (Minimizing deadtime reduces losses in the diodes).
When you have parallel banks of MOSFETs making a single switch then you have to match within each bank so that you get equal current sharing during the switching transient (if one device hogs all the current it can be a problem). Perhaps that's what you are thinking of?