Our Jenkins environment has a bunch of templates that share a common AMI, but use difference instance sizes and snapshot images. As an example:

MachineA:
ami-12345
Root Device size: 60GB
R4.4xlarge
Instance cap: 3

MachineB:
ami-12345
Root Device size: 20GB
t2.medium
User-Data installs docker
Instance Cap: 1

We hit a common occurrence where if we have three instances of A running, EC2Plugin will not provision machine B when a job requests for that machine label. Upon inspection of the source code, available instances to launch are computed by the number of running AMIs, not by the template name.

Upon reading the help blurb, the "Instance Cap" field on the configure page is carefully worded so that it is limited by the AMI and "associated configurations". From a user interface point of view, I would expect the instance cap to be set per-template (per-description)