bearfoil, my mind skipped a beat on the TCAS. I meant ACARS. It's antenna has a wide beamwidth and can steer. How well it tracks on turns with steep banks I don't know. I figure it would take nearly a full aircraft roll well past 60 degrees bank for a turn to the right to cut off communications until near the end of a 180 degree turn. A left hand turn would be shadowed sooner for a satellite to their East along the Equator. That might go out with a 45 degree bank at a rude guess. That's if the antenna steering gets aiding from the navigation systems' turn and bank sensors as well as the position data. Otherwise a rapid bank could bring the beam too far off the satellite at about 30 degrees bank for a turn to the left.
Once the situation had pickled silence on 121.5 indicates a very busy pair of pilots. And I am so communications oriented I tend to scratch my head excessively (no, no dandruff {^_-}) over the lack of affirmative attempts to get into communications with Dakar. That just "feels wrong." I probably dwell on it too much. About the only explanation IF being in communications was seen as an imperative, is two incapacitated pilots at the controls. And that's too much of a stretch.
If they DID have communications they could have cried emergency and given their "exact" position as the altimeter unwound. MAYBE somebody could have been fished out of the drink. (Indications are that would not have happened. But, a priori the pilots would not know they could not plant it without breaking it.)
mm43, there is a good argument that things started to go sour right around the time of that confirmed SELCAL test at 01:35:43z. As soon as the end of that transmission something bad may have happened OR they changed to DAKAR and never bothered to reestablish communications.
Regarding frequencies, limited is quite true. At least enough are available that what is audible from here near ONT indicates frequencies are not "shared" within rather large radii. And they do pick frequencies that suit the required communications distance. 5MHz is a band that can perform on extremely long hauls as well as close in. At night it would open up for "thousands" of miles, even in the solar minimum. Although during the recent solar minimum that might have been very near the maximum usable frequency for F2 bounces and that only to very great distances. That's a period I heard 3.5 MHz go numb some nights - usually well after 2400 local time. I am presuming the frequency choice was made intelligent. That's well understood by the people involved.
Finally someone mention something about nominal comms procedures. I am surprised it's appropriate to have a plane on a modestly busy traffic route out of communications for any significant period of time. (I'm only half surprised that there are not two HF receivers with appropriate TR switching so that both ATLANTICO and DAKAR could be monitored at the same time. That might be deemed "expensive". But for receive only not much is needed except the T/R switching to protect the second receiver. And mumbling off topic from my experience the military Automatic Link Establishment protocol could easily be bent into automatic link establishment for aircraft. It's a digital mode that sounds like polite turkey gobbles. It's designed to find optimum frequencies for communications dynamically in real time..)
Machinbird, released from the bottom and bubbling up to the surface in a sharp edged pattern as shown seems improbable to me. I'd expect more dispersion from the random direction changes bubbles experience on their way up. So I'd want to see some information about kerosene releases at that depth to be certain. The length of that "tail" is about right to my intuition. The width seems too narrow. I don't see a scale, for comparison, on the lower picture. Nor does it indicate the depth of release or the hydrocarbon fraction leaked. Heavy bunker oil will certainly remain more coherent on the way up than a lighter fraction like kerosene.
{^_^}