Diversity is often associated with the functional stability of ecological communities from microbes to macroorganisms. Understanding how diversity responds to environmental perturbations and the consequences of this relationship for ecosystem function are thus central challenges in microbial ecology. Unimodal diversity-disturbance relationships, in which maximum diversity occurs at intermediate levels of disturbance, have been predicted for ecosystems where life history tradeoffs separate organisms along a disturbance gradient. However, empirical support for such peaked relationships in macrosystems is mixed, and few studies have explored these relationships in microbial systems. Here we use complex microbial microcosm communities to systematically determine diversity-disturbance relationships over a range of disturbance regimes. We observed a reproducible switch between community states, which gave rise to transient diversity maxima when community states were forced to mix. Communities showed reduced compositional stability when diversity was highest. To further explore these dynamics, we formulated a simple model that reveals specific regimes under which diversity maxima are stable. Together, our results show how both unimodal and non-unimodal diversity-disturbance relationships can be observed as a system switches between two distinct microbial community states; this process likely occurs across a wide range of spatially and temporally heterogeneous microbial ecosystems. The diversity of microbial communities is linked to the functioning and stability of ecosystems. As humanity continues to impact ecosystems worldwide, and as diet and disease perturb our own commensal microbial communities, the ability to predict how microbial diversity will respond to disturbance is of critical importance. Using microbial microcosm experiments, we find that community diversity responds to different disturbance regimes in a reproducible and predictable way. Maximum diversity occurs when two communities, each suited to different environmental conditions, are mixed due to disturbance. This maximum diversity is transient except under specific regimes. Using a simple mathematical model, we show that transient unimodality is likely a common feature of microbial diversity-disturbance relationships in fluctuating environments.