I have experienced the same thing, and have come to the conclusion that it must be some bug in the cache invalidation event system when a page is published by a scheduled job.
That seems like an major bug! Havent found it in the buggy bug list.
Is there an event (scheduler publish) I could override to send my own cache invalidation?
We've the similar envionment and facing issues with composer global blocks - it does not invalidate cache instantly and sometimes take between 10-20 minutes .
I also have some pages that published with a scheduler havent notice that if does not clear cache instantly but i will check this now . Any one know if anything wrong with composer global blocks not updating instantly ?
Note : we are updating composer global block in edit mode (not from admin mode) so I expect this to clear cache instantly.
I have a weird problem...
Running CMS 6R2 Enterprise with load balancing and events configured for TCP.
If I publish a page manually the cache gets updated on the front ends, thus the cache invalidation events seems to work. everything is just fine.
BUT!
If I choose to delay a publish using the scheduler (set to run every minute) then the published page wont show up until as long as eight minutes later.
I looked in the logs for the delayed scheduler and ther is no entry for the publish (0) pages published, but the published page is showing up in the edit mode (which is not cached at all) and aprox 8 minutes or so it has found its way to the load balancing nodes as well.
Any Ideas?