The length limits come from two requirements: signal quality, and signal delays. A 1m cable will introduce about 5ns of delay (speed of light in circuits is a lower than in vacuum!), so if you are running at 1GHz, there are five symbols in transit on that 1m cable. It this particular case it probably does not matter (it's more important that the signal paths are of exactly the same length), but this is the sort of consideration that most people won't think about.
The standard (at least what's open to the public: http://mipi.org/specifications/physical-layer) only mentions length considerations for M-PHY ("Distance: optimized for short interconnect (<10 cm) but extendable to a meter with good quality interconnect or even further with optical converters and optical waveguides"); D-PHY is a lot simpler, so it should have no problem with longer lengths as far as signal delays are concerned.
As @joan said, a long cable makes you more susceptible to interference from others. It also makes you more likely to interfere with other equipment, your neighbor's old analog TV, or maybe another pi on the next shelf with equally long cables. Hence my original admonition of keeping things as short and as straight as possible.
If you are thinking of shielding the flat cable with aluminum foil, you may or may not get a working cable: you will be adding a lot of capacitance, changing the transmission line characteristics.
I realize this is probably more detail than you needed, but someone may find it useful in the future :)