This is a perf issue typically seen for larger deployments (advanced zone, user accounts ~4k, lot of guest networks). The issue happens when deployVM is called without 'networkids' parameter. In this case a search is done to identify the default guest network to use for deploying VM. The search logic doesn't scale up for larger deployments resulting in delay in returning async job id.
Tests results showed that for ~4k user accounts and similar number of guest networks it takes approx. 15 secs. to return async job id.