>> It'll take a few minutes but I'll try and get the information to you. A summary
>> is:
>>
>> Process 1:
>> - writing 50 rows/second, 1 row/transaction.
>> - every so often delete 100 rows
>>
>> Process 2:
>> - running ANALYZE VERBOSE and pg_total_relation_size every second
>>
>
> You'll probably have to vacuum pg_statistics as well then.
>
>
I presume because of the frequent ANALYZEs? In my real-world application
I won't be running ANALYZE manually like this. I am only using it as a
quick hack to get a picture of the dead rows in the various tables so I
could get a feel for what was happening, and it seems that this is what
is causing a conflict with the autovacuum...