Assistance with an out of shared memory error

Hi Postgres group,

I'm reaching out for some help with an "ERROR: out of shared memory. HINT: You might need to increase max_locks_per_transaction" issue:

I've got a PL/pgSQL function (regenerate_gateway_last_seen, attached) that loops through all partitions of 2 tables ("Measure" and "ValuelessMeasure", schema attached) selecting data from each into another table ("GatewayLastSeenNew"). Occasionally the function runs to completion, but usually it errors with the message copied below. I'd appreciate any advice to help understand why I'm seeing this and if increasing max_locks_per_transaction, changing another configuration value, or changing how the function works would improve reliability.

At the moment I have three theories as to why this might be happening:
1. Machine is out of memory - this seems unlikely because shared_buffers is '5400MB' and I don't see the machine take more of the available swap when I run the function.

2. max_locks_per_transaction is being hit in the transaction - this also seems unlikely because max_locks_per_transaction is set to the default value of 64, but there are ~8000 partitions to consider and I expect the resulting GatewayLastSeenNew table to have thousands of rows. If I was taking locks that would contribute towards that ceiling of 64 I'd expect this to fail every time, instead of failing often but not always as I observe.

3. The max size of the lock table is being exceeded - my understanding is that the lock table has room for max_locks_per_transaction * (max_connections + max_prepared_transactions) locks, which would be 64 * (200 + 0) = 12800 for my current configuration. I used 'SELECT COUNT(*) FROM PG_LOCKS' while the function was running and observe values as high as 21758, so if this is the issue it seems like I might not be estimating the max size of the lock table correctly or only specific locks contribute to that.

Thanks,
Gautam