* This is part of the bits needed to facilitate multiple/incremental
  resource updates on I/O servers.

* Instead the update is tagged to be synchronized later.
* Secondary updates to resources are not yet handled by the PIO code
  it's meant for.

* Also use no-fail [mc]alloc wrappers.

* Also use no-fail allocations.

* The code in question might otherwise access a deleted resource.

* Also add code to test for it with valgrind.