There is no reason you cannot run pega on heterogenous nodes. However, please be aware that this is very rare in reality (espcially in production env), please perform extensive tests if you have to use this configuration. In GCS, we run this configuration all the time, but they are not production ready systems.
Nothing comes to mind. But again, we do not run any real use cases in these systems either. What is the reason that you need to run two disparate environment within the same cluster (i.e., same database)?
We run out of IBM System P server resources (there are no new LPARs available and they are bloody expensive) and would like to switch to x86 completely. That process might need some time. Meanwhile, we would like to add x86 node to make our customer happy ASAP.