Checks running in suggestion mode currently work in a way whose performance cannot scale to support a large number of suggestions on long documents. It is clear we need performance improvements, else response times would become absurdly slow.

We should determine SLOs for edit suggestions, likely incorporating some of the following metrics:

• Response time (synchronous and asynchronous)
• Number of simultaneous checks supported
• Number of simultaneous actions
• Memory usage
• Response takes too long (read: latency)
• Cache miss (value is not in the cache, so we ask Headless VE to compute on-the-fly)

This will feed into the work of determining what performance improvements to implement.