Taking code coverage of processes which runs many PL/I programs

Our request is to capture a code coverage data of over 10,000 programs running on a IMS FP region without restarting a IMS FP region.

Our online modules running on IMS FP are managed by SAIL framework, a IBM Japan asset programs. There, basically LE enclave is not recreated on the next program running, so only one enclave is used during FP region running.
With all programs running in debug mode on one FP region connecting to a headless mode collector, a code coverage data of each program run is not written to a file, and all code coverage data of all program run are kept in e-private storage of FP region.
Therefore, as the number of program executions increases, the amount of free e-private space will decrease, and then user program can not run at last.

As a result of our evaluation project last year, after around 300 to 500 programs run on one FP region, e-private was full and FP region must start again. We had to run around 10,000 programs for one day regression test, so they have to restart FP region many times. Therefore, we had to give up all tests and had to test only a limited number of program.

We understand following issues.
a) the code coverage results are all lumped together as a single "run"
b) the data would accumulate until the region was finally stopped.

Regarding above issues, we have the following thoughts.
a) The purpose of taking code coverage is for quality verification of PL/I migration. Therefore, it is more important to know the instructions executed in the entire test than to know the instructions executed by each transaction.
b) We understand that we have a limited number, but we want to run as many programs as possible.