Techniques are provided for using a middlebox to composite displays from multiple hosted virtualized applications (HVAs) for transmission to a client endpoint device. The middlebox receives one or more HVA displays via a virtual desktop interface (VDI) protocol, each HVA display from an individual HVA. The middlebox renders and composites the HVA displays together into a hosted virtual desktop (HVD) display, and sends the HVD display to the client endpoint device via a VDI protocol. The client endpoint device is therefore able to display a composite image of multiple HVA displays even if it is a zero client endpoint lacking the capability to perform compositing itself. In some examples, the middlebox reduces computational load by reducing the HVD frame rate, so that it is able to maintain HVD functionality during times of high system activity.