huge shared_blocks_hit one select but manually run very fast

Hi,
   we have a simple select .... from table where ... (that mache the index)
, table has 80million rows.  when many application sessions run the query
and at the same time some other sessions doing insert into ... this table.
from pg_stat_statements, shared_blks_hit show 31652 / per call.   we see
very high cpu almost 100% cpu during application workload test, and high
LWLock BufferMapping waiting for these querys.  But manually run the sql
show only 2148 shared_blks_hit/ per call.  this is a simple sql, from
pg_profile we did see it use same index scan as manually running.  What
could be possible reason leading so big difference with shared_blks_hit ?
 PGv14.8

Thanks,

James