Memory spike while building 1.29 binaries crashes staging runs

[puerco]

What happened:

When running builds in parallel, memory consumption jumps around the time go-runner is being built. This causes the staging process to exit as memory is exhausted.

What you expected to happen:

Builds should finish successfully

How to reproduce it (as minimally and precisely as possible):

Run a staging run of the release-1.29 branch. Or those with access can check out run c56008f6 in the GCB console.

Anything else we need to know?:

We mitigated the crashing stage run by switching to serial builds. This is a temporary measure, however as it causes the builds to run slower.