The push task currently depends on all other tasks in the variant, meaning that a single test failure will prevent the push task from running.

Since the push task is the one that propagates the -latest binaries that downstream teams use for testing, flaky tests and increased batch times means these binaries may not be updated in a long time.

By limiting the tasks the push tasks depend on to a small subset we'll be able to generate binaries for testing more often, which will help with downstream integration testing. As a bonus, we'll be exercising the push task itself more regularly, which will allow us to catch bugs when we make changes to it (which will help with release management and package generation).