In our production setup we get timeout errors to confluence log when trying to add Jira Calendar for certain project. This project has 2 custom date-time issue fields defined in it.

Diagnosed Jira team calendar integration locally and found that when custom fields are used, JIRA plugin makes one query for each issue:

SELECT ID, ISSUE, CUSTOMFIELD, PARENTKEY, STRINGVALUE, NUMBERVALUE, TEXTVALUE, DATEVALUE, VALUETYPE FROM PUBLIC.customfieldvalue WHERE ISSUE='10100'

Our theory is that when project contains thousands of issues this would result in a considerable lag due to accumulated network latency.

Debugging Jira plugin code revealed that the problem lies in
DefaultDateCustomFieldDateProvider line 48:

Object customFieldValue = issue.getCustomFieldValue(customField);

which results in an SQL call.

The idea for a fix: make one query for all the project's issues and then access required values from memory. I have not found API support for this, so possibly it would imply working with SQL directly.