Generalized Stochastic Petri Nets (GSPNs) are an established tool for representing and analyzing concurrency, timing, synchronization, precedence, and priority in processes. GSPNs emerged in the 1980s as the de facto standard for modeling stochastic processes using Petri nets, supported by tools such as GreatSPN. However, traditional applications of this technology assume that GSPNs are created manually. Given the widespread availability of event data in information systems, this seems sub-optimal. This explains the uptake of process mining, which starts from event data instead of manually created process models. There are dozens of techniques to discover basic (i.e., non-stochastic) Petri nets given an event log. However, there is an increasing interest in not just discovering control flow, but also learning the stochastic behavior based on event data. Therefore, we take GSPNs as the target representation for process discovery. Since there are numerous techniques to discover the control flow, we focus on the extensions provided by GSPNs. These include priorities, blocking, probabilities, and rates. In this paper, we sketch a concrete approach to discover probabilities and rates from event data. This is done by translating to GSPNs to Markov chains to which parameter synthesis is applied. Since priorities and blocking can be added without limiting GSPN-based performance analysis, we advocate the development of control-flow discovery techniques incorporating these features. Having GSPNs learned from event data, we can support more forward-looking forms of process mining.