We had an issue with parsing queue filling recently. Our oversized event profile is to blame. To address, I increased our queue size across the board to 100 MB from 50 MB. Is there any reason not to go higher? 1 GB? Given an indexer with unlimited RAM, is there a point of diminishing returns when it comes to queue sizing? Do the different queues have different PoDRs?
We've run pretty high queue sizes in the order of multiple gigabytes - but if you've got regular massive queues instead of short-term spikes you're probably focusing on the wrong problem. Ensure that you've sized your servers appropriately for the load, and tune the processing path to make sure you're not unnecessarily holding up data.