I'm not familiar with the shared memory handling, but could we not allocate just enough shared memory to fit the data we're going to write instead of the fixed 8k? It's not that we cannot know the length of the resulting plan text in advance.
the shared memory cannot be reused - (released) :(, so allocating enough memory is not effective. More - in this moment it has not sense. Shared memory queue can do almost all work.
A-ha, I've discovered the shared memory message queue facility and I see how we can use it.
But do we really need the slots mechanism? Would it not be OK to just let the LWLock do the sequencing of concurrent requests? Given that we only going to use one message queue per cluster, there's not much concurrency you can gain by introducing slots I believe.