dcintegration-1.3.4-1.16 has been observed blocking the server thread for over 10 seconds in de.erdbeerbaerlp.dcintegration.Discord.getWebhook() / net.dv8tion.jda.api.requests.RestAction.complete(). This was presumably due to its blocking getWebhook implementation paired with Discord having a period of particularly slow API responses. Another occurrence blocked the server for 36 seconds, but I didn't capture a trace for it. These interactions should never block the server thread.

From what I can tell at a quick glance, JDA offers RestAction.submit instead of .complete to obtain a CompletableFuture for the webhook. CompletableFuture.thenAccept and similar methods can then be used to execute the desired action (.send+.close) when the original operation has completed - without waiting/blocking the caller. You want to finish the CompletableFuture chain with .whenComplete or .exceptionally to print any exceptions, otherwise they get lost.

VisualVM compatible nps file showing the issue:
http://files.player.to/tmp/spike-10486-16338923416.nps
The file shows a capture for a single tick only. Server thread is waiting 10.4s on the completion of a JDA future, which presumably corresponds to the request execution on JDA RateLimit-Worker 1.

This is probably the number one forgecraft killer right now. It seems this happens regularly, and causes many second lockups of the serverthread, sometimes resulting in watchdog kills :(