Too many dynamic shared memory segments error at select

[pashkinelfe]

at 9c34d00

Reproduction from #333.
Insert tuples into 10000 partitions (2000 sec, maybe less)
In parallel backend run:

select count(*) from data;
ERROR:  too many dynamic shared memory segments
select count(*) from data_1;
 count
-------
  6954
select count(*) from data_2;
 count
-------
  6846

[pashkinelfe]

At doing select it was doing a parallel scan of partitions.
Parallel worker stack for 60s: (200 samples/s)

[pashkinelfe]

error is reproduced with max_connections = 300. With max_connections = 3000 there is no error

[pashkinelfe]

Plan (fails at max_connections = 300, succeds at max_connections = 3000)

 postgres=# explain select count(*) from data;
                                          QUERY PLAN
----------------------------------------------------------------------------------------------
 Finalize Aggregate  (cost=166875.22..166875.23 rows=1 width=8)
   ->  Gather  (cost=166875.00..166875.21 rows=2 width=8)
         Workers Planned: 2
         ->  Partial Aggregate  (cost=165875.00..165875.01 rows=1 width=8)
               ->  Parallel Append  (cost=0.00..158250.00 rows=3050000 width=0)
                     ->  Parallel Seq Scan on data_1  (cost=0.00..14.30 rows=430 width=0)
                     ->  Parallel Seq Scan on data_2  (cost=0.00..14.30 rows=430 width=0)
                     ->  Parallel Seq Scan on data_3  (cost=0.00..14.30 rows=430 width=0)
                     ->  Parallel Seq Scan on data_4  (cost=0.00..14.30 rows=430 width=0)
                     ->  Parallel Seq Scan on data_5  (cost=0.00..14.30 rows=430 width=0)
                     ->  Parallel Seq Scan on data_6  (cost=0.00..14.30 rows=430 width=0)
                     ->  Parallel Seq Scan on data_7  (cost=0.00..14.30 rows=430 width=0)
                     ->  Parallel Seq Scan on data_8  (cost=0.00..14.30 rows=430 width=0)
                     ->  Parallel Seq Scan on data_9  (cost=0.00..14.30 rows=430 width=0)
                     ->  Parallel Seq Scan on data_10  (cost=0.00..14.30 rows=430 width=0)
<10000 table scans continued>

Plan with

[pashkinelfe]

Rechecked at fix 32d4e91 with max_parallel_workers_per_gather = 2 (default) and 15.
Error doesn't appear anymore.