they don't see a sufficient return on the costs of development.
This is, more or less, the same idea as the "Zero Marginal Productivity" attempt at explaining the employment effects surrounding the global financial crisis (see http://marginalrevolution.com/marginalrevolution/2010/07/zero-marginal-product-workers.html for a handy starting point, but it's a widely discussed idea).
In the case of GSS, it's clearly fairly wide of the mark; Ghost was and remains one of the two best investments ever made by Symantec (the most well-known of the two being Norton Antivirus), and despite all the incomprehensibly bad management actions taken during Symantec's stewardship of the product it was still very profitable when it was shut down in 2009.
[ As a side note, our team made a point of internal numerical transparency; it took a lot of digging to get the data (lots of internal barriers, generally, as well as the VP who cancelled Ghost in 2004 being dead-set against us having any metrics we could use to appeal his actions), but the cost and revenue numbers for GSS (or Ghost generally) were made available within the team so we could all see how well or otherwise we were doing. Note that during the relatively brief time Symantec half-heartedly tried to use Net Promotor Score, our NPS numbers were at the top of the scale too; our NPS numbers alone were the kind of thing that - if you believe in NPS, anyway - were pretty valuable just for the halo effect on Symantec's brand. ]
At the point our studio was closed in 2009, the mean revenue per employee across all the development and Q&A staff (since product Q&A was done on-site and it's really part of R&D expenditure) was around US$4m per employee, and remember the reinvestment rate was under 3%. We weren't just profitable, we were a cash cow, and despite the low reinvestment rate we had some useful things on the table for future releases (such as being able to clone *to* a system while it was still running, for example).
A more convincing explanation is this:
Other than general cost reduction due to the GFC, Symantec had - for quite some time - a general program of reducing real-estate expenditures by closing sites, especially sites outside North America but even including NA ones, and these frequently were done even though they would destroy products. This program of site closures existed prior to the GFC, but the GFC caused it to be massively accelerated.
pre-GFC PCAnywhere, for instance, was doing about US$20mil of revenue, still profitable (not Ghost-profitable, but enough since it was a "portfolio" product), and was about to release - literally months away - a cloud offering to compete with GoToMyPC et al as part of the Symantec Protection Network. However the real-estate people wanted to close the site in Newport News, VA in which the PCA team was housed (they occupied only a small part of the facility). The team was basically told to sell up - during a housing-market crash in Newport News due to the military force realignments underway - and move to the facility in Utah or leave. So, development on the PCA product was stopped, the launch of their new online product was pulled, they were all laid off, and PCA's maintenance moved to India.
This has been an ongoing thing in Symantec; usually there's enough pushback to see that it mostly only affects dead products or really underperforming teams, but this affects lots of sites. Pretty much the only Symantec site outside NA or India that was able to escape the closure program is the former Altiris one in Talinn, Estonia - they do good work, but their survival really owes the fact that their cost structure is even better than New Zealand's (where during 2004/2005, we were picking up products like WinFAX to maintain at our site because our R&D costs were so low and we had spare developers since Ghost was cancelled during those years).
Anyway, the bottom line is that due to strange forces like real-estate portfolio management, Symantec has a number of times in the past (and probably will again in the future) kill off profitable and not at all ZMP products for reasons that really have nothing to do with the product itself.