Sample size limit in unbounded sequences

I just started an evaluation of Connext (rti_connext_dds-6.1.1) and I am exploring the capabilities of unbounded sequences, but I ran into a problem when my sample reaches a certain size.  I modified the Hello World "getting started" example as follows: I have an IDL file that contains the following structs...

struct A {
  double a1;
  double a2;
  double a3;
};
struct B {
  uint32 b1;
  uint32 b2;
  uint32 b3;
};
struct C {
  sequence<A> c1;
  sequence<B> c2;
};

Note that the two fields of type 'C' are intentionally unbounded. I ran rtiddsgen on the IDL file using the following switches...

  -language c++11
  -platform x64Linux4gcc7.3.0
  -create makefiles
  -create typefiles
  -unboundedSupport
  -alwaysUseStdVector

and I used a USER_QOS_PROFILES.xml file containing the following elements (shortened and formatted for this post)...

<name>dds.data_writer.history.memory_manager.fast_pool.pool_buffer_max_size</name><value>4096</value>
<name>dds.data_reader.history.memory_manager.fast_pool.pool_buffer_max_size</name><value>4096</value>

In my publishing code I keep the sizes of the 'c1' and 'c2' fields the same for simplicity.

In my code I publish samples of type 'C' and successfully read the samples when my sequences are below a specific size, around 1800 or so, but when I make my sequences larger than that size, my subscriber no longer reads any samples.  I'm pretty new to Connext and to DDS in general. Am I missing some settings that would allow larger sequences in my example? Or is there perhaps an underlying limit imposed by the RTI libraries for unbounded sequences?

Thank you.