Got a weird suggestion about a troublesome Tomcat installation running in a subsystem

We have a customer box on which a Tomcat server (and more importantly, BIRT reports running in a webapp context within it) is running much slower than it should be.

We have it running in a subsystem, spec'd out thusly:

Pool Storage Activity
ID Size (K) Level
1 *BASE
2 4194304 400
Seq Nbr Program Library
9999 QCMD QSYS
Routing entry sequence number . . . . . . . : 9999
Program . . . . . . . . . . . . . . . . . . : QCMD
Library . . . . . . . . . . . . . . . . . : QSYS
Class . . . . . . . . . . . . . . . . . . . : QBATCH
Library . . . . . . . . . . . . . . . . . : QSYS
Maximum active routing steps . . . . . . . : *NOMAX
Pool identifier . . . . . . . . . . . . . . : 2
Compare value . . . . . . . . . . . . . . . : *ANY

Compare start position . . . . . . . . . . :
Thread resources affinity:
Group . . . . . . . . . . . . . . . . . . : *SYSVAL
Level . . . . . . . . . . . . . . . . . . :
Resources affinity group . . . . . . . . . : *NO

(Note that this is *double* the usual size of the private memory pools we'v set up.)

My understanding is that this gives preference to the private memory pool, but will fall back on *BASE as needed. Am I correct?

At any rate, we just got the following suggestion from one of their people:

why not consider changing the subsystem you set up to not have its
own pool, but run it out of the *BASE pool. That way, it will grab
more memory if needed.

That seems to go directly against what little I know about subsystems and memory pools. Is there any merit in it? Or is there anything else that could be made better?

--
JHHL