Assuming the control panel could be updated to allow USB cache to be used then a couple of changes to the web interface could perhaps be considered:
1) The 50 limit cut-off line could be changed from a hard limit to a soft one. Classic users with no USB cache file would only have access to the widgets above the line. Those with a cache file would gain access to all of them.
2) Rather than making it a limit on the number of widgets, make it a limit on the size of the widgets. To keep it simple, give each widget a score from 1-20 based on how large it is, with the 400k licking dog as a 20. With 2MB of cache available on a Classic, you could have up to five 400k widgets, giving a maximum of 100 points - but you would probably set the limit lower (e.g. 80) to allow for some overhead. It's up to you whether you have 80 x 1 point widgets, 4 x 20 points, or a combination in between. Adding a USB cache would increase your points allowance (or remove the limit entirely).
This is obviously more work to implement, but you do gain a few benefits:
* By using a non-linear scale to allocate the points you can subtly encourage users to go for more lightweight widgets
* You're setting a real limit based on actual sizes, rather than hoping that "average widget size" corresponds to "average size of widgets used".
* The biggest bandwidth users are the ones most penalised, and most encouraged to add a USB cache.
* Low bandwidth users are no longer subsidising bigger users.
* It could actually be sold as an *increase* in the allowance, since you could now have up to 80 widgets (if you choose wisely), rather than the hard limit of 50.
Because it would still be possible to have multiple channels, each with 80 points to allocate, there's no real reduction in functionality for anyone, but from the server perspective you would know that each channel's widgets are entirely cacheable and therefore only have to be served once for each channel change, at most.
Ideally I'd like to see the cost of the subscription come down a little in the medium term, provided it doesn't affect the long term sustainability. Reducing the bandwidth costs is likely a necessary step towards that goal. Encouraging the use of a USB cache, and mitigating the effect on the servers for those Classic users who don't have such a cache, would seem to be a step in that direction.