The PCI SCC states the following in their penetration testing guidance:
Per PCI DSS Requirements 11.3.1 and 11.3.2, penetration testing must be performed at least annually and after any significant change—for example, infrastructure or application upgrade or modification—or new system component installations. What is deemed “significant” is highly dependent an entity’s riskassessment process and on the configuration of a given environment. Because of this variability, a significant change is not prescribed by PCI DSS. If the change could impact the security of the network or allow access to cardholder data, it may be considered significant by the entity. Penetration testing of significant changes is performed to ensure that controls assumed to be in place are still working effectively after the upgrade or modification
Therefore it is up to individual organisations to define what constitutes a significant change in their documentation. However, as the penetration test should include security assessments on any in-scope applications present, an organisation may have specified that new features in these applications constitute a significant change.
If that is the case, is it allowable under PCI DSS to only "pen test" the changed applications, or does PCI dictate that the whole infrastructure be retested too, even though these individual servers may not have changed? Yes, in any ecosystem there's the law of unintended consequences, however if it can reasonably be implied that the changes won't affect anything else does testing have to be fully repeated?
Of course if it is allowable, this should be documented as such that this is the line that the organisation is taking. Also, as there is an annual penetration test of everything in-scope, this would cover any new vulnerabilities that may affect the rest of the network.