Deadlock in SAP RFC destination

Using IS SAP adapter, we are experiencing some stange behaviour.
First, we had ~1000 IDocs in the tRFC queue (if I get terminology confused here, forgive me - I can spell SAP, that’s about all) and we initiated a background job to send these to IS. The job hung and I had to reboot the IS server - IS would not respond to a restart command.
Then we started manually re-submitting the queued jobs. This worked, but my colleague may have got over-zealous and appears to have submitted faster than SAP/IS could handle. The flow of IDocs stopped again. The listener showed green so I stopped and started it again. Everything started flowing again.
These IDocs are quite small (9Kb) and I am wondering if SAP acquires a lock on the RFC destination when it is about to place an IDoc there and if IS acquires a lock when it is about to remove an IDoc. If so, is there any chance of getting into deadlock when the volumes are high, but the message sizes are small and therefore moving in and out rapidly ?
Has anybody else come across this behaviour ?
If so, what was the solution ?