Having Jenkins instance with a vast number of rundeck jobs Jenkins startup time increases dramatically. With 1500+ rundeck jobs on very strong AWS instance Jenkins starts up over 30 minutes. That number of jobs can be easy to achieve having automatically generated (e.g. with Jenkins JobDSL) deployment pipelines for multiple products deployed in multiple countries.
After the analyze it turned out that rundeck-plugin makes a hit to Rundeck about job details for every runcked job in Jenkins (in RundeckJobProjectLinkerAction to display details on a build page). It can take some time, especially if Rundeck is configured to use H2 DB.
In many cases number of distinct jobs in Rundeck is very limited (e.g. several) - they are prameterized. Having it cached reduce number all calls dramatically. In our case for 1500+ rundeck jobs (several distinct rundeck jobs) the startup time was reduced from over 30 minutes to less than 30 seconds .
I will created PR with proposed implementation as it could be probably useful for others.