It's all coming back now...I do remember you describing this history!
With the marginal articulation, he would have to use human scribing, and, preferably, oral assessment. I would affirm that recommendation. The key really is the huge difference between his written and oral output. The exercise you described ("astronaut", etc.) is exactly like one of the tasks on the CELF-5, which is one of the core language assessments given by SLPs. If he did well on that task (which he should have, and which the absence of language therapy Rx suggests), but struggled with the written version of it for multiple hours, that is a really stark difference between oral and written language skills. If it is not the case that his ability to complete this task orally is substantially better than his ability to complete it in writing (which would be evident on a CELF-5), then that may suggest an actual language expression disorder, which would indicate speech/language therapy, not just for articulation.
And of course, I'm not a neurologist nor an occupational therapist, but given his history of both severe speech articulation and handwriting weaknesses, I'm a little surprised there was no further investigation regarding motor planning skills, as that is one of the deficits that would be involved in both of those presentations.
Processing speed and working memory really aren't all that related to each other per se. They're grouped together because they both primarily affect efficiency (hence output), rather than abstract reasoning. Also, they are both sensitive to a number of other common concerns, such as challenges with attention, emotional interference, low motivation, fatigue, etc.
Both processing speed subtests do involve pencil work, but one of them is just slashes, while the other is letter-like forms. Notice he did equally Average on both. But the letter-like forms are unfamiliar to all students (hypothetically), so no one really gets to use automaticity skills on them, with the exception of the minority who memorize them in the brief timespan of the subtest. His original VMI scores were extremely high in visual perception, still quite high in integration, and on the lower end of Average in coordination. These were, of course, untimed scores.
His BD score probably reflects his visual perceptual/spatial reasoning more than his speed or fine-motor performance, mainly due to his age. His chronological peers aren't getting enough items correct to separate by speed (three years from now, this score may begin to reflect effects of speed more than accuracy, as the number of items correct obtained by the median student rises), so the fact that he correctly completed many designs, albeit slowly, is more important than the lack of bonus points for speed.
Though I too am a little puzzled by the lower Visual Puzzle score. It does have time limits, though. Also, not true that every timed subtest is below 14. Arithmetic is timed, as are Word Recognition Fluency and Associational Fluency, which are the equivalent of 14s.
Not as puzzled by Matrix Reasoning, as it is very close to Figure Weights. I'm more interested in how he is this strong in math, when his most math-associated cognitive tasks are just average/high average (other than the actual math task, Arithmetic).