Former Theranos lab director Adam Rosendorff was the second employee who testified in Elizabeth Holmes’ trial who kept his work emails. Forwarding work emails to a personal account can violate a non-disclosure agreement, which Rosendorff signed when he joined the company. But, like Surekha Gangakhedkar before him, he was worried he’d be blamed for the company’s problems.

He was right to worry: he’s one of the people Elizabeth Holmes’ defense is trying to blame.

In opening arguments of US v Elizabeth Holmes, the defense agreed there were problems in Theranos’ lab. But Lance Wade, Homes’ attorney, said that problems in the clinical lab were ultimately the responsibility of the lab director. And besides, that lab director reported to Holmes’ co-defendant, Sunny Balwani, who is being tried separately. They face 10 counts of wire fraud and two counts of conspiracy to commit wire fraud.

Today’s testimony was about bad Theranos tests. A lot of new emails were introduced, showing Holmes was aware of the company’s problems, and was even actively trying to manage the situation. Several times in those emails, Rosendorff tried to get Theranos labs to run FDA-approved tests instead of the ones Theranos developed. But maybe even more telling were the emails that Rosendorff was excluded from.

Rosendorff didn’t mince words when he took the witness stand. He quit because of many things, he said. “One was the unwillingness of management to perform proficiency testing as required by law, I felt pressured to vouch for tests I did not have confidence in.” That wasn’t all. He came to believe “the company was more about PR and fundraising than patient care,” he said.

When he first joined Theranos in April 2013, Rosendorff thought Theranos might be the next Apple. But the sheen wore off the company quickly — and he began looking for other jobs by that summer. In August, before Theranos’ launch in patients, the clinical lab was “anxious” about testing real people, he said; the pace around the launch “extremely rushed and hurried.”

On August 31, 2013, Elizabeth Holmes sent a 1AM email asking for an update on how many tests had completed validation. An employee wrote back that none of them had. The launch was scheduled for September 9th, 2013, and the schedule came from Holmes and Balwani, Rosendorff said.

Increasingly worried, he tried to stall it. He was “raising alarm bells.” In an email a little more than a week before the scheduled launch, Rosendorff wrote to Holmes and Balwani raising concerns about some of the tests. He also thought the lab needed more staff with better training.

He needed “a few more weeks to sort through these medical and logistical issues,” he wrote in the email. That would mean delaying bringing Theranos’ tests to patients. He even met with Holmes about his concerns, because he felt Balwani wasn’t taking him seriously.

During that meeting Rosendorff described Holmes as “very nervous.” “She was not her usual composed self. She was trembling a little bit,” he said.

Despite Rosendorff’s concerns, the launch went ahead as planned.

Some of his testimony corroborated what previous Theranos employees had said. He echoed Erika Cheung’s concerns about removing outliers from Theranos data. As both Cheung and Gangakhedkar testified, the Edison machines frequently failed quality control.

While a quarter of Theranos’ devices routinely failed quality control, failures of commercially-available tech was far less frequent. Theranos devices “failed so frequently it raised doubts in my mind regarding the accuracy of the tests themselves,” Rosendorff said. He emailed Balwani about the failure rates, who wrote back, “Adam. This is not the case.”

Normally lab directors deal with doctors who have concerns about lab results — this was part of Rosendorff’s job at his previous employer, the University of Pittsburgh. But Theranos installed Holmes’ brother, Christian, as a liaison for complaints. Rosendorff said he was pressured to come up with explanations about bad results that didn’t call into question the tests themselves.

But the tests were bad. In the case of a key pregnancy hormone — one that we heard testimony about from a patient who was wrongly told she was miscarrying — Rosendorff sent an all-caps email to his staff in May 2014 telling them that all future tests were to be run on an FDA-approved device.

This didn’t happen.

In fact, there was a June 2014 email chain where Christian Holmes wrote his sister, CCing Balwani, that the test was “causing serious complaints and patient issues.” Additionally, the lab was “a complete mess.”

Holmes wrote back that she’d connect with her brother, then added, “Sunny / this is already handled.”

There are a few notable things about the exchange. First, it demolishes the idea that Holmes could have been misled about the state of the clinical lab and its tests. Second, it shows her clearly taking charge of the situation, even telling Balwani to stand down. But third, it’s remarkable for who’s missing from the chain: Rosendorff, who may not have even been aware that his instruction to use FDA-cleared devices only was ignored.

The first time he saw those emails was in court, he said. It was not the only set of emails about patient complaints we saw where Rosendorff wasn’t included, either.

In February of 2014, Rosendorff also tried to convince the company to stop running its HDL (“good”) cholesterol test on Theranos devices. He sent an email to Balwani and added Holmes, because he thought Balwani wasn’t listening to him. “I got a lot of pushback” for suggesting the lab use approved devices, Rosendorff said.

That pushback came from Holmes, Balwani, and VP Daniel Young, he testified. And though Young didn’t have a medical degree, he often offered suggestions about what was “really” wrong with tests, Rosendorff said. Through his testimony, Rosendorff generally talked about Young in a tone that suggested he loathed the VP.

The problems in the lab were widespread, Rosendorff testified, pointing to tests of bicarbonate, and an email where he pointed out that two-thirds of patients were reading below the normal range. “The test has lost any diagnostic value,” he emailed in September 2014

But Theranos wasn’t honest with physicians or providers about this. Instead, customer service representatives were instructed to say that the results “were not reported due to temporary unavailability.”

Another email from a healthcare provider said it didn’t realize Theranos’ methods weren’t FDA-approved, and complained about testosterone and blood sugar tests. “I am not sure what to do with these lab results????” the email read.

Part of the problem was that all the validation studies had been done on venous blood, Rosendorff testified. See, to get enough fingerprick blood to run all the studies, you’d need to prick every finger on the patient — so as a matter of convenience, venous blood was drawn. But if you draw blood from fingerpricks, you’re far more likely to destroy red blood cells in the process, leading to faulty readings.

Among the bad tests was one for sodium, potassium, and chloride, called ISE. Young, who was a statistician, not a doctor, suggested the faulty results for this test was because of the red blood cells breaking apart, and urged a visual examination of the blood to confirm.

This was bad for several reasons. First of all, the FDA-approved machines detected this immediately, not after a weird reading. Second, evidence of broken red blood cells wasn’t always obvious to the eye, Rosendorff testified.

This bad assay had been used on patients since September 2013. In June 2014, there were still problems with the tests, and Young’s suggested fixes, including a “bias-correction” method, still weren’t working. So Theranos voided tests that fell above and below accepted values.

That was a problem, too. We saw an email from a doctor dated October 27th, 2014, addressed to a customer service rep. The patient had low sodium levels — which was why the test had been ordered. A customer service rep emailed Rosendorff saying, “Is it possible that that was the value? If she comes in again, and the value is still critical low, will it be voided again?”

Rosendorff forwarded this email to Balwani and Holmes, writing, “I’m not sure of a clinical value of a sodium assay in which the only time we can report it is when it is not critical and the very situations that require accurate measurement and reporting of abnormal sodium results are voided.”

Calcium testing, too, was bad. “Are there any findings we can relay about high trends in calcium reporting?” Christian Holmes wrote to Young and Balwani. “Seems to be by far the majority of questions we get about accuracy now.”

He then forwarded this to his sister. “It’s pretty obvious we have issues with calcium, potassium, and sodium, specifically,” he wrote. “Obviously I can’t tell them we are wrong, but they continue to send patients to Quest [another blood-testing company] after we report high, and the results continue to come back normal.”

Rosendorff wasn’t included on this chain, and took issue with “specifically.” There were lots of problems with other tests, he said.

“I didn’t believe at this point that Theranos could solve these problems,” Rosendorff testified. He began avoiding speaking to doctors, and told one or two physicians that he didn’t trust the results.

All of this is bad for Holmes — not only is she aware of the problems, she’s calling the shots. But her defense hinted at what the cross-examination of Rosendorff may bring. One of his subsequent employers was also investigated by the federal authorities, as well as another company that was subject to a justice department probe. (We found this out before the jury was allowed in the courtroom.)

It’s not clear that Rosendorff’s post-Theranos employment will be deemed relevant to the current case — the jury may hear nothing about this. But the defense will certainly have some work to do. Though Rosendorff’s testimony was much more technical than that of board member James Mattis yesterday, it was possibly even worse for Holmes. If she’s trying to convince the jury that any fraud was mostly Balwani’s fault, she’s going to have to explain the ample evidence she was giving marching orders.