the signer uses locks in GS [1] to manage jobs across a fleet of systems to provide reliability and automatic recovery from failures.  but some low level flakes in GS itself sometimes breaks us (see issue b/112110347 for details).

buildbucket v2 API [2] might be able to provide a lot of the same requirements and hopefully none of the flakes.  so we should investigate using that to manage the task queue.  it might also include some features we don't have today (e.g. metrics) as well as replace more features (e.g. our e-mail alert system is pretty rudimentary).

it seems WebRTC moved to buildbucket v2 [3] which might have parallels to our needs.

we should also compare this to pubsub in issue 923945 so we can make an informed decision as to the different trade-offs.

[1] https://docs.google.com/document/d/1Yqv_G4QlMYOzJt9h_OU0564OAApPqekkTfFJI9cdJ-M/edit
[2] http://go/buildbucket-design
[3] https://docs.google.com/document/d/1XcFsvdgins8DAGme4Twzx9FGKDznfrPAeI_sFxudbzw/edit