Perpendicular to X-ray beam In order to optimize your chances of collecting such data, I recommend the following:
Spot shape and size is determined by many things, but one primary factor is the illuminated crystal volume. Therefore, reducing the size of the beam to 0.1mm makes the spot smaller. Also, using smaller crystals may be useful, if still they diffract reasonably well. Note that intense low-resolution reflections are often "larger" than typical spots because of detector hardware limitations (spot "bloom"). You may need to do a second low-resolution pass if this becomes a big factor in the loss of low-resolution data. Overflows are the other reason to do low resolution passes but this is not dependent on cell dimensions.
In order to measure data reliably, there must be at least one pixel of background between adjacent spots. Combined with spot size, this imposes a finite closest-approach for the detector to the crystal. Although you canlie about the spot size in Denzo during integration, this always compromises data quality. Reducing the angular width of the oscillation is not going to help here since spot spacing is simply a factor of unit cell size, wavelength and detector geometry.
Spot spacing is linear with crystal-detector distance, so to double the spot spacing you need to double the crystal-detector distance. This means you tend to get into the low-resolution range rather quickly on smaller detectors (e.g. the MAR 165 at X9A/X3). In cases of really large unit cells you will have to use beamlines that have bigger (ADSC Quantum4 or Quantum315) detectors such as X29 or X25.
Effective spot spacing is maximised when the long cell axis is in the plane of the detector (i.e. perpendicular to the direct beam in most configurations). If it is not, you get a projection phenomenon that is fairly simply a function of the inclination to the beam:
effective-spacing = max-spacing * sin(angle-to-beam)
Perpendicular to X-ray beam
45 degrees to X-ray beam
15 degrees to X-ray beam
When the long axis is parallel to the beam, you're essentially looking down the long axis and all the spots get thrown out due to spots piling up on each other (spatial "overlaps"). No cute data collection strategies will rescue you in this orientation.
During data collection you'll start to lose data due to overlaps unless you arrange the long axis so that it is nearly parallel to the spindle axis. If you don't do this, the long axis moves away from the plane of the detector as you rotate the crystal during data collection, and you end up with the projection phenomenon (i.e. you look down the long axis and lose all the data).
In many cases the high symmetry axis is often the long axis. In this case in particular, it is vital that you put the long axis along the spindle since this is also the most efficient data collection method from the point of view of symmetry.
So, just mount your crystal so that the long axis is along the spindle axis and collect data. Knowing the relationship between your crystal morphology and unit cell direction is vital if you to reliably mount crystals in this orientation.
On beamlines with kappa goniostats you can sometimes use the kappa axis to reposition the crystal such that it is now along the spindle axis. Kappa offsets the phi axis from the omega axis. Usually at kappa=0 the phi and omega axes are colinear. At least some kappa axes have a restricted range of motion such that you cannot orient the pin arbitrarily with respect to the rotation axis (often omega, sometimes phi). On X29, the kappa axis is not aligned very well with omega, so you need to recenter the crystal if you move kappa, and it . Using the kappa axis is very much easier than using extended arc goniometer heads, but you should still pay attention to the allowed physical limits of motion for kappa - especially when the detector is relatively close to the crystal (e.g. 200mm or closer). The X29 kappa axis cannot reorient the pin so it sits vertically (kappa=180 puts the pin at approximately 45 degrees to omega). Beamlines with half or full circle goniostats are relatively rare but allow a fuller range of motion than single axis or kappa axis goniostats. If you truly cannot mount your crystal advantageously you either have to use the Supper head or these more flexible (and bulky) goniostats.
As far as the "extended arc" method is concerned: to use this method you have to be EXCEPTIONALLY careful and check everything at least twice, otherwise the extended Supper arc can crash into the beamstop, the collimator and various other critical hardware components in the beamline . You will trash the beamline and incur the wrath of the beamline staff and the next users.
Do not use this method unless you are very comfortable with the workings of the beamline - get a more experienced person to do it for you.
The following discussion assumes you are using a Supper extended arc goniometer head at a synchrotron (horizontal phi/omega axis).
Step #1: find the long axis.
Mount the crystal in the normal way. Collect very short exposures
every 30 degrees or so until you find the location of the long axis.
Find the best orientation such that the long axis is mostly vertical
on the detector. Make a note of the angle (and make note of
the appearance of the xtal/loop in the microscope at this point).
Then, find the orientation of the goniometer head that would allow you
to mount the extended arc in the vertical plane. Note this angle too.
Step #2: reorient the crystal on the gonimeter head
With the two orientations from step one, you need to manually
move the crystal on the goniometer head until the long axis lies
in the plane of the extended arc on the goniometer head. I.E. you
want the crystal to have it's long axis on the detector when the
goniometer head is aligned such that the extended arc is vertical.
Since the MAR base is just a one-axis goniostat you need to
do it via moving the crystal on the goniometer head rather than using
beamline hardware (goniostat). Note that you don't need the long arc
in place to do this. Be sure to test the re-orientation by taking
a short image with the goniometer head in the correct orientation.
Note the angle that the long axis makes on the detector surface.
Step #3: attach the extended arc and determine the physical limits
Attach the extended arc. Make a very careful note of the
physical limits that you can operate this within, such that you do
not bend the backstop or crash into the collimator or anything else.
Check that you know this range before proceeding, or you will break
something during data collection. Also, the extended arc itself will
block the diffracted beams at some angles.
Step #4: move the crystal up the extended arc
Slide the sled and attached crystal up the extended arc until it's in
a position where you would predict that the long axis is more-or-less
horizontal. You will need to recenter the crystal, which on some
beamlines can be a tricky task with a vertical crystal orientation.
Take a quick image at this angle to make sure you are right.
If all is well, then you will not need to re-orient the crystal.
Otherwise tweak the orientation until you have achieved your goal.
At this point, if you've got a firm mental image of the relationship
between your long axis and your crystal, you're ready to go.
Step #5: collect data
Be paranoid. Check the acceptible ranges of angles through which
you can safely rotate the crystal. Then go ahead and collect the
data, making sure to continuously monitor the goniometer head
orientation. One mistake and you may break and/or disable the
beamline to the considerable inconvenience of yourself and other users. There is
no room for error here. In difficult cases it may be necessary
to reorient the crystal on the goniostat multiple times in order to
collect data, since the goniometer head can only safely rotate through
a small solid angle with the extended arc attached. If your space group
is P1 or C2 or P21 this method is just not going to work
for you. It probably won't work in orthorhombic space groups either.
Find another beamline and a (much) bigger detector.