The load-archive merge caused Bad Gateway problems

[gfr10598]

This became apparent when looking at etl pipeline in staging. Discovered after basing a new sandbox branch on master with the merged load-archive code.

I have NOT rolled back or fixed the code. I HAVE reverted the versions running in both sandbox and staging, but the problem will recur if we redeploy this code.

[gfr10598]

Another hint at possible cause - profile showed a large amount of time spent in futex? Also, prometheus was getting very occasional snapshots, so the servers were there, but usually unresponsive.

[pboothe]

Obsolete.