You can of course front-run manually, depending on the degree of automation in the systems one is front-running, but front-runners will generally want to automate. And to scale the scheme for a realtime system of FTX's volume, Alameda would certainly want to automate it.
I wasn't following the case closely enough, but did the prosecution have access to Alameda source code, or was it only FTX? As long as Alameda had access to FTX order records, any front-running code would likely have resided in Alameda repositories.
This blog post [0] contains screenshots of git commits that were used as evidence in the trial. It only shows commits to FTX's source code, so I don't know if the prosecution also had access to Alameda's source code.
Given how obviously incriminating those snippets are, I don't think SBF really tried to hide his crimes. I'm not sure then why they would architecture FTX such there are no signs of front-running in their code.
Personally, my guess is that either A) the prosecution found evidence of front-running but thought that the case was air-tight enough already and they didn't want to confuse the jurors or B) Alameda didn't actually do any front-running.
[0]: https://newsletter.mollywhite.net/p/the-fraud-was-in-the-cod...