The geothermal reservoir at Soultz-sous-Forets is a valuable natural laboratory for understanding the mechanisms of microearthquakes generated during stimulations and circulation tests. An ongoing effort currently exists regarding the retrieval of mechanisms aimed to indicate the type of fracturing of the rock massif. As a default, a moment tensor description has been applied. Nevertheless, the retrieval of the mode of fracturing still remains ambiguous. Recent studies indicate a prevailing shear slip but, rarely, a non-shear pattern has also been observed. The moment tensor, used today as a universal tool for descriptions of the mechanism, captures general balanced dipole sources. However, in the case of small-scale earthquakes, the moment tensor need not always be reliably determined. In an effort to fit the data, there may be notable non-shear components caused by the low quality of input data. Constraining the source model to directly determine a simpler one is convenient for describing the physical phenomena expected for a particular focus. An opening of new fractures can be described, to a first approximation, by a tensile crack, optionally combined with a shear slip. Such an alternative model is called a shear-tensile crack (STC) source model. The combination is practical, and can be used to both identify events that reflect purely mode-I (tensile) failure and to determine the dilation angle of the fracture undergoing shear. The latter is particularly important in enhanced geothermal system reservoirs such as Soultz, where shear-related dilation is believed to be the primary mechanism underpinning permeability creation during stimulation injections. We performed a synthetic case study by simulating seismic data as recorded by the actual seismic array installed at Soultz-sous-Forets. Synthetic P and S amplitudes for several shear-tensile source models were inverted for several types of station coverage. The analysis explored how results were influenced by mislocation, mismodeling, and noise contamination of data. In most cases, the orientation of the mechanism was well resolved. Determination of shear vs. non-shear content within the mechanism was more difficult. From all of the factors influencing resolution that we explored, the quality of the monitoring system (the number of stations and their distribution with respect to the focus) and noise contamination were of the highest impact. The STC source model yielded considerably less spurious non-shear fracture components than the moment tensor. From the bulk of the seismicity recorded during the stimulation in 2003, we concentrated on the first phase of the injection when only a single borehole at the site was stimulated. We processed thirteen small earthquakes with magnitudes larger than 1.4 that were not treated in previous studies. We determined that their source mechanisms were dominantly pure shear slips on pre-existing faults, just as the earthquakes investigated earlier were. The results were also in agreement with the stress pattern from in situ measurements.