Conclusions

• A detailed comparison of several measures of the image quality shows that the median peak FI and peak SNR are good estimates, being far more reliable than even weighted means.
• Both uv-plots and simulations show that the observations most compromised by a shift from C to CS are snapshots, for which CS degrades the image qualities by up to a factor of two. Longer syntheses certainly show holes at intermediate spacings with CS which are not present with C, but this does not seem to significantly degrade the resulting images.
• A single-antenna version of CS would basically solve the intermediate spacing problem, particularly if the current `three-antenna' rule were changed to a `two-antenna' rule for CS configuration. This would give somewhat worse uv-coverage than two-antenna versions of CS. More importantly, moving only one antenna would leave the short spacing advantage of CS alarmingly vulnerable to the loss of a single antenna. Special treatment of that single antenna (i.e. automatic call-out whenever it goes down) would help but not solve this problem, and would involve many complications in terms of late-night summons, maintenance headaches, and the like. A more fool-proof solution would involve the automatic granting of D configuration observing time to projects which lose this single CS antenna.
• Of the possible CS configurations, the 1997 version (EW12$\rightarrow$EW3) seems comparable to the other obvious option (EW10$\rightarrow$EW3) if one wants two short-spacing antennas. If one such antenna is considered sufficient, it should probably be transferred from an intermediate pad on the north arm (N10 or N12) to N1.