So what's going on with that gene therapy? A paper in 2000 by Yazawa et al. used a mouse model for an initial feasibility study. Though it doesn't appear that they actually used the method for tumor reduction, they did show that B. longum into mice with lung tumors could only be found in tumor tissue after 168 hours. That's presumably because the bacteria require an anaerobic (or, at least a hypoxic) environment to survice. Ideally, this means that B. longum bearing some kind of anti-tumor factor could be injected into or near a tumor with no pathogenic effect on any other tissues. B. longum can be killed off using common antibiotics; Yazawa et al. used ampicillin, though they only tried it in vitro. I'd be worried about long-term use with immunocompromised patients, though the exact anti-tumor material in play may be another critical factor.
The same research group was apparently still working on the idea as of 2010, when they published this review. A 2013 study by a different group looks like it had some success in using Bifidobacterium longum subsp. infantis as part of a method for treating bladder cancer in a mouse model.* This 2014 study just coated their Bifidobacterium in selenium, an elemental micronutrient which may have anti-tumor properties.
*I don't have access to the article so I'm not sure how well the method worked. The authors claim their treatment "exhibited the highest level of apoptosis" compared to controls so that could just mean they had a statistically-significant but limited effect on tumors. Cancer therapy isn't really my field so I'll give them the benefit of the doubt.